linux/drivers/misc/uacce/uacce.c

652 lines
15 KiB
C
Raw Permalink Normal View History

uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
// SPDX-License-Identifier: GPL-2.0-or-later
#include <linux/compat.h>
#include <linux/dma-mapping.h>
#include <linux/iommu.h>
#include <linux/module.h>
#include <linux/poll.h>
#include <linux/slab.h>
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
#include <linux/uacce.h>
static dev_t uacce_devt;
static DEFINE_XARRAY_ALLOC(uacce_xa);
static const struct class uacce_class = {
.name = UACCE_NAME,
};
/*
* If the parent driver or the device disappears, the queue state is invalid and
* ops are not usable anymore.
*/
static bool uacce_queue_is_valid(struct uacce_queue *q)
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
{
return q->state == UACCE_Q_INIT || q->state == UACCE_Q_STARTED;
}
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
static int uacce_start_queue(struct uacce_queue *q)
{
int ret;
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
if (q->state != UACCE_Q_INIT)
return -EINVAL;
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
if (q->uacce->ops->start_queue) {
ret = q->uacce->ops->start_queue(q);
if (ret < 0)
return ret;
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
}
q->state = UACCE_Q_STARTED;
return 0;
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
}
static int uacce_put_queue(struct uacce_queue *q)
{
struct uacce_device *uacce = q->uacce;
if ((q->state == UACCE_Q_STARTED) && uacce->ops->stop_queue)
uacce->ops->stop_queue(q);
if ((q->state == UACCE_Q_INIT || q->state == UACCE_Q_STARTED) &&
uacce->ops->put_queue)
uacce->ops->put_queue(q);
q->state = UACCE_Q_ZOMBIE;
return 0;
}
static long uacce_fops_unl_ioctl(struct file *filep,
unsigned int cmd, unsigned long arg)
{
struct uacce_queue *q = filep->private_data;
struct uacce_device *uacce = q->uacce;
long ret = -ENXIO;
/*
* uacce->ops->ioctl() may take the mmap_lock when copying arg to/from
* user. Avoid a circular lock dependency with uacce_fops_mmap(), which
* gets called with mmap_lock held, by taking uacce->mutex instead of
* q->mutex. Doing this in uacce_fops_mmap() is not possible because
* uacce_fops_open() calls iommu_sva_bind_device(), which takes
* mmap_lock, while holding uacce->mutex.
*/
mutex_lock(&uacce->mutex);
if (!uacce_queue_is_valid(q))
goto out_unlock;
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
switch (cmd) {
case UACCE_CMD_START_Q:
ret = uacce_start_queue(q);
break;
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
case UACCE_CMD_PUT_Q:
ret = uacce_put_queue(q);
break;
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
default:
if (uacce->ops->ioctl)
ret = uacce->ops->ioctl(q, cmd, arg);
else
ret = -EINVAL;
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
}
out_unlock:
mutex_unlock(&uacce->mutex);
return ret;
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
}
#ifdef CONFIG_COMPAT
static long uacce_fops_compat_ioctl(struct file *filep,
unsigned int cmd, unsigned long arg)
{
arg = (unsigned long)compat_ptr(arg);
return uacce_fops_unl_ioctl(filep, cmd, arg);
}
#endif
static int uacce_bind_queue(struct uacce_device *uacce, struct uacce_queue *q)
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
{
u32 pasid;
struct iommu_sva *handle;
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
if (!(uacce->flags & UACCE_DEV_SVA))
return 0;
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
iommu: Remove SVM_FLAG_SUPERVISOR_MODE support The current kernel DMA with PASID support is based on the SVA with a flag SVM_FLAG_SUPERVISOR_MODE. The IOMMU driver binds the kernel memory address space to a PASID of the device. The device driver programs the device with kernel virtual address (KVA) for DMA access. There have been security and functional issues with this approach: - The lack of IOTLB synchronization upon kernel page table updates. (vmalloc, module/BPF loading, CONFIG_DEBUG_PAGEALLOC etc.) - Other than slight more protection, using kernel virtual address (KVA) has little advantage over physical address. There are also no use cases yet where DMA engines need kernel virtual addresses for in-kernel DMA. This removes SVM_FLAG_SUPERVISOR_MODE support from the IOMMU interface. The device drivers are suggested to handle kernel DMA with PASID through the kernel DMA APIs. The drvdata parameter in iommu_sva_bind_device() and all callbacks is not needed anymore. Cleanup them as well. Link: https://lore.kernel.org/linux-iommu/20210511194726.GP1002214@nvidia.com/ Signed-off-by: Jacob Pan <jacob.jun.pan@linux.intel.com> Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com> Reviewed-by: Jason Gunthorpe <jgg@nvidia.com> Reviewed-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Fenghua Yu <fenghua.yu@intel.com> Tested-by: Zhangfei Gao <zhangfei.gao@linaro.org> Tested-by: Tony Zhu <tony.zhu@intel.com> Link: https://lore.kernel.org/r/20221031005917.45690-4-baolu.lu@linux.intel.com Signed-off-by: Joerg Roedel <jroedel@suse.de>
2022-10-31 00:59:07 +00:00
handle = iommu_sva_bind_device(uacce->parent, current->mm);
if (IS_ERR(handle))
return PTR_ERR(handle);
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
pasid = iommu_sva_get_pasid(handle);
if (pasid == IOMMU_PASID_INVALID) {
iommu_sva_unbind_device(handle);
return -ENODEV;
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
}
q->handle = handle;
q->pasid = pasid;
return 0;
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
}
static void uacce_unbind_queue(struct uacce_queue *q)
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
{
if (!q->handle)
return;
iommu_sva_unbind_device(q->handle);
q->handle = NULL;
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
}
static int uacce_fops_open(struct inode *inode, struct file *filep)
{
struct uacce_device *uacce;
struct uacce_queue *q;
int ret;
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
uacce = xa_load(&uacce_xa, iminor(inode));
if (!uacce)
return -ENODEV;
q = kzalloc(sizeof(struct uacce_queue), GFP_KERNEL);
if (!q)
return -ENOMEM;
mutex_lock(&uacce->mutex);
if (!uacce->parent) {
ret = -EINVAL;
goto out_with_mem;
}
ret = uacce_bind_queue(uacce, q);
if (ret)
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
goto out_with_mem;
q->uacce = uacce;
if (uacce->ops->get_queue) {
ret = uacce->ops->get_queue(uacce, q->pasid, q);
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
if (ret < 0)
goto out_with_bond;
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
}
init_waitqueue_head(&q->wait);
filep->private_data = q;
q->state = UACCE_Q_INIT;
q->mapping = filep->f_mapping;
mutex_init(&q->mutex);
list_add(&q->list, &uacce->queues);
mutex_unlock(&uacce->mutex);
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
return 0;
out_with_bond:
uacce_unbind_queue(q);
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
out_with_mem:
kfree(q);
mutex_unlock(&uacce->mutex);
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
return ret;
}
static int uacce_fops_release(struct inode *inode, struct file *filep)
{
struct uacce_queue *q = filep->private_data;
struct uacce_device *uacce = q->uacce;
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
mutex_lock(&uacce->mutex);
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
uacce_put_queue(q);
uacce_unbind_queue(q);
list_del(&q->list);
mutex_unlock(&uacce->mutex);
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
kfree(q);
return 0;
}
static void uacce_vma_close(struct vm_area_struct *vma)
{
struct uacce_queue *q = vma->vm_private_data;
if (vma->vm_pgoff < UACCE_MAX_REGION) {
struct uacce_qfile_region *qfr = q->qfrs[vma->vm_pgoff];
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
mutex_lock(&q->mutex);
q->qfrs[vma->vm_pgoff] = NULL;
mutex_unlock(&q->mutex);
kfree(qfr);
}
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
}
static const struct vm_operations_struct uacce_vm_ops = {
.close = uacce_vma_close,
};
static int uacce_fops_mmap(struct file *filep, struct vm_area_struct *vma)
{
struct uacce_queue *q = filep->private_data;
struct uacce_device *uacce = q->uacce;
struct uacce_qfile_region *qfr;
enum uacce_qfrt type = UACCE_MAX_REGION;
int ret = 0;
if (vma->vm_pgoff < UACCE_MAX_REGION)
type = vma->vm_pgoff;
else
return -EINVAL;
qfr = kzalloc(sizeof(*qfr), GFP_KERNEL);
if (!qfr)
return -ENOMEM;
mm: replace vma->vm_flags direct modifications with modifier calls Replace direct modifications to vma->vm_flags with calls to modifier functions to be able to track flag changes and to keep vma locking correctness. [akpm@linux-foundation.org: fix drivers/misc/open-dice.c, per Hyeonggon Yoo] Link: https://lkml.kernel.org/r/20230126193752.297968-5-surenb@google.com Signed-off-by: Suren Baghdasaryan <surenb@google.com> Acked-by: Michal Hocko <mhocko@suse.com> Acked-by: Mel Gorman <mgorman@techsingularity.net> Acked-by: Mike Rapoport (IBM) <rppt@kernel.org> Acked-by: Sebastian Reichel <sebastian.reichel@collabora.com> Reviewed-by: Liam R. Howlett <Liam.Howlett@Oracle.com> Reviewed-by: Hyeonggon Yoo <42.hyeyoo@gmail.com> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arjun Roy <arjunroy@google.com> Cc: Axel Rasmussen <axelrasmussen@google.com> Cc: David Hildenbrand <david@redhat.com> Cc: David Howells <dhowells@redhat.com> Cc: Davidlohr Bueso <dave@stgolabs.net> Cc: David Rientjes <rientjes@google.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Greg Thelen <gthelen@google.com> Cc: Hugh Dickins <hughd@google.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Jann Horn <jannh@google.com> Cc: Joel Fernandes <joelaf@google.com> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Kent Overstreet <kent.overstreet@linux.dev> Cc: Laurent Dufour <ldufour@linux.ibm.com> Cc: Lorenzo Stoakes <lstoakes@gmail.com> Cc: Matthew Wilcox <willy@infradead.org> Cc: Minchan Kim <minchan@google.com> Cc: Paul E. McKenney <paulmck@kernel.org> Cc: Peter Oskolkov <posk@google.com> Cc: Peter Xu <peterx@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Punit Agrawal <punit.agrawal@bytedance.com> Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Shakeel Butt <shakeelb@google.com> Cc: Soheil Hassas Yeganeh <soheil@google.com> Cc: Song Liu <songliubraving@fb.com> Cc: Vlastimil Babka <vbabka@suse.cz> Cc: Will Deacon <will@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-01-26 19:37:49 +00:00
vm_flags_set(vma, VM_DONTCOPY | VM_DONTEXPAND | VM_WIPEONFORK);
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
vma->vm_ops = &uacce_vm_ops;
vma->vm_private_data = q;
qfr->type = type;
mutex_lock(&q->mutex);
if (!uacce_queue_is_valid(q)) {
ret = -ENXIO;
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
goto out_with_lock;
}
if (q->qfrs[type]) {
ret = -EEXIST;
goto out_with_lock;
}
switch (type) {
case UACCE_QFRT_MMIO:
case UACCE_QFRT_DUS:
if (!uacce->ops->mmap) {
ret = -EINVAL;
goto out_with_lock;
}
ret = uacce->ops->mmap(q, vma, qfr);
if (ret)
goto out_with_lock;
break;
default:
ret = -EINVAL;
goto out_with_lock;
}
q->qfrs[type] = qfr;
mutex_unlock(&q->mutex);
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
return ret;
out_with_lock:
mutex_unlock(&q->mutex);
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
kfree(qfr);
return ret;
}
static __poll_t uacce_fops_poll(struct file *file, poll_table *wait)
{
struct uacce_queue *q = file->private_data;
struct uacce_device *uacce = q->uacce;
__poll_t ret = 0;
mutex_lock(&q->mutex);
if (!uacce_queue_is_valid(q))
goto out_unlock;
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
poll_wait(file, &q->wait, wait);
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
if (uacce->ops->is_q_updated && uacce->ops->is_q_updated(q))
ret = EPOLLIN | EPOLLRDNORM;
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
out_unlock:
mutex_unlock(&q->mutex);
return ret;
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
}
static const struct file_operations uacce_fops = {
.owner = THIS_MODULE,
.open = uacce_fops_open,
.release = uacce_fops_release,
.unlocked_ioctl = uacce_fops_unl_ioctl,
#ifdef CONFIG_COMPAT
.compat_ioctl = uacce_fops_compat_ioctl,
#endif
.mmap = uacce_fops_mmap,
.poll = uacce_fops_poll,
};
#define to_uacce_device(dev) container_of(dev, struct uacce_device, dev)
static ssize_t api_show(struct device *dev,
struct device_attribute *attr, char *buf)
{
struct uacce_device *uacce = to_uacce_device(dev);
return sysfs_emit(buf, "%s\n", uacce->api_ver);
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
}
static ssize_t flags_show(struct device *dev,
struct device_attribute *attr, char *buf)
{
struct uacce_device *uacce = to_uacce_device(dev);
return sysfs_emit(buf, "%u\n", uacce->flags);
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
}
static ssize_t available_instances_show(struct device *dev,
struct device_attribute *attr,
char *buf)
{
struct uacce_device *uacce = to_uacce_device(dev);
if (!uacce->ops->get_available_instances)
return -ENODEV;
return sysfs_emit(buf, "%d\n",
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
uacce->ops->get_available_instances(uacce));
}
static ssize_t algorithms_show(struct device *dev,
struct device_attribute *attr, char *buf)
{
struct uacce_device *uacce = to_uacce_device(dev);
return sysfs_emit(buf, "%s\n", uacce->algs);
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
}
static ssize_t region_mmio_size_show(struct device *dev,
struct device_attribute *attr, char *buf)
{
struct uacce_device *uacce = to_uacce_device(dev);
return sysfs_emit(buf, "%lu\n",
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
uacce->qf_pg_num[UACCE_QFRT_MMIO] << PAGE_SHIFT);
}
static ssize_t region_dus_size_show(struct device *dev,
struct device_attribute *attr, char *buf)
{
struct uacce_device *uacce = to_uacce_device(dev);
return sysfs_emit(buf, "%lu\n",
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
uacce->qf_pg_num[UACCE_QFRT_DUS] << PAGE_SHIFT);
}
static ssize_t isolate_show(struct device *dev,
struct device_attribute *attr, char *buf)
{
struct uacce_device *uacce = to_uacce_device(dev);
return sysfs_emit(buf, "%d\n", uacce->ops->get_isolate_state(uacce));
}
static ssize_t isolate_strategy_show(struct device *dev, struct device_attribute *attr, char *buf)
{
struct uacce_device *uacce = to_uacce_device(dev);
u32 val;
val = uacce->ops->isolate_err_threshold_read(uacce);
return sysfs_emit(buf, "%u\n", val);
}
static ssize_t isolate_strategy_store(struct device *dev, struct device_attribute *attr,
const char *buf, size_t count)
{
struct uacce_device *uacce = to_uacce_device(dev);
unsigned long val;
int ret;
if (kstrtoul(buf, 0, &val) < 0)
return -EINVAL;
if (val > UACCE_MAX_ERR_THRESHOLD)
return -EINVAL;
ret = uacce->ops->isolate_err_threshold_write(uacce, val);
if (ret)
return ret;
return count;
}
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
static DEVICE_ATTR_RO(api);
static DEVICE_ATTR_RO(flags);
static DEVICE_ATTR_RO(available_instances);
static DEVICE_ATTR_RO(algorithms);
static DEVICE_ATTR_RO(region_mmio_size);
static DEVICE_ATTR_RO(region_dus_size);
static DEVICE_ATTR_RO(isolate);
static DEVICE_ATTR_RW(isolate_strategy);
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
static struct attribute *uacce_dev_attrs[] = {
&dev_attr_api.attr,
&dev_attr_flags.attr,
&dev_attr_available_instances.attr,
&dev_attr_algorithms.attr,
&dev_attr_region_mmio_size.attr,
&dev_attr_region_dus_size.attr,
&dev_attr_isolate.attr,
&dev_attr_isolate_strategy.attr,
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
NULL,
};
static umode_t uacce_dev_is_visible(struct kobject *kobj,
struct attribute *attr, int n)
{
struct device *dev = kobj_to_dev(kobj);
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
struct uacce_device *uacce = to_uacce_device(dev);
if (((attr == &dev_attr_region_mmio_size.attr) &&
(!uacce->qf_pg_num[UACCE_QFRT_MMIO])) ||
((attr == &dev_attr_region_dus_size.attr) &&
(!uacce->qf_pg_num[UACCE_QFRT_DUS])))
return 0;
if (attr == &dev_attr_isolate_strategy.attr &&
(!uacce->ops->isolate_err_threshold_read &&
!uacce->ops->isolate_err_threshold_write))
return 0;
if (attr == &dev_attr_isolate.attr && !uacce->ops->get_isolate_state)
return 0;
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
return attr->mode;
}
static struct attribute_group uacce_dev_group = {
.is_visible = uacce_dev_is_visible,
.attrs = uacce_dev_attrs,
};
__ATTRIBUTE_GROUPS(uacce_dev);
static void uacce_release(struct device *dev)
{
struct uacce_device *uacce = to_uacce_device(dev);
kfree(uacce);
}
static unsigned int uacce_enable_sva(struct device *parent, unsigned int flags)
{
int ret;
if (!(flags & UACCE_DEV_SVA))
return flags;
flags &= ~UACCE_DEV_SVA;
ret = iommu_dev_enable_feature(parent, IOMMU_DEV_FEAT_IOPF);
if (ret) {
dev_err(parent, "failed to enable IOPF feature! ret = %pe\n", ERR_PTR(ret));
return flags;
}
ret = iommu_dev_enable_feature(parent, IOMMU_DEV_FEAT_SVA);
if (ret) {
dev_err(parent, "failed to enable SVA feature! ret = %pe\n", ERR_PTR(ret));
iommu_dev_disable_feature(parent, IOMMU_DEV_FEAT_IOPF);
return flags;
}
return flags | UACCE_DEV_SVA;
}
static void uacce_disable_sva(struct uacce_device *uacce)
{
if (!(uacce->flags & UACCE_DEV_SVA))
return;
iommu_dev_disable_feature(uacce->parent, IOMMU_DEV_FEAT_SVA);
iommu_dev_disable_feature(uacce->parent, IOMMU_DEV_FEAT_IOPF);
}
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
/**
* uacce_alloc() - alloc an accelerator
* @parent: pointer of uacce parent device
* @interface: pointer of uacce_interface for register
*
* Returns uacce pointer if success and ERR_PTR if not
* Need check returned negotiated uacce->flags
*/
struct uacce_device *uacce_alloc(struct device *parent,
struct uacce_interface *interface)
{
unsigned int flags = interface->flags;
struct uacce_device *uacce;
int ret;
uacce = kzalloc(sizeof(struct uacce_device), GFP_KERNEL);
if (!uacce)
return ERR_PTR(-ENOMEM);
flags = uacce_enable_sva(parent, flags);
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
uacce->parent = parent;
uacce->flags = flags;
uacce->ops = interface->ops;
ret = xa_alloc(&uacce_xa, &uacce->dev_id, uacce, xa_limit_32b,
GFP_KERNEL);
if (ret < 0)
goto err_with_uacce;
INIT_LIST_HEAD(&uacce->queues);
mutex_init(&uacce->mutex);
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
device_initialize(&uacce->dev);
uacce->dev.devt = MKDEV(MAJOR(uacce_devt), uacce->dev_id);
uacce->dev.class = &uacce_class;
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
uacce->dev.groups = uacce_dev_groups;
uacce->dev.parent = uacce->parent;
uacce->dev.release = uacce_release;
dev_set_name(&uacce->dev, "%s-%d", interface->name, uacce->dev_id);
return uacce;
err_with_uacce:
uacce_disable_sva(uacce);
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
kfree(uacce);
return ERR_PTR(ret);
}
EXPORT_SYMBOL_GPL(uacce_alloc);
/**
* uacce_register() - add the accelerator to cdev and export to user space
* @uacce: The initialized uacce device
*
* Return 0 if register succeeded, or an error.
*/
int uacce_register(struct uacce_device *uacce)
{
if (!uacce)
return -ENODEV;
uacce->cdev = cdev_alloc();
if (!uacce->cdev)
return -ENOMEM;
uacce->cdev->ops = &uacce_fops;
uacce->cdev->owner = THIS_MODULE;
return cdev_device_add(uacce->cdev, &uacce->dev);
}
EXPORT_SYMBOL_GPL(uacce_register);
/**
* uacce_remove() - remove the accelerator
* @uacce: the accelerator to remove
*/
void uacce_remove(struct uacce_device *uacce)
{
struct uacce_queue *q, *next_q;
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
if (!uacce)
return;
/*
* uacce_fops_open() may be running concurrently, even after we remove
* the cdev. Holding uacce->mutex ensures that open() does not obtain a
* removed uacce device.
*/
mutex_lock(&uacce->mutex);
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
/* ensure no open queue remains */
list_for_each_entry_safe(q, next_q, &uacce->queues, list) {
/*
* Taking q->mutex ensures that fops do not use the defunct
* uacce->ops after the queue is disabled.
*/
mutex_lock(&q->mutex);
uacce_put_queue(q);
mutex_unlock(&q->mutex);
uacce_unbind_queue(q);
/*
* unmap remaining mapping from user space, preventing user still
* access the mmaped area while parent device is already removed
*/
unmap_mapping_range(q->mapping, 0, 0, 1);
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
}
/* disable sva now since no opened queues */
uacce_disable_sva(uacce);
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
if (uacce->cdev)
cdev_device_del(uacce->cdev, &uacce->dev);
xa_erase(&uacce_xa, uacce->dev_id);
/*
* uacce exists as long as there are open fds, but ops will be freed
* now. Ensure that bugs cause NULL deref rather than use-after-free.
*/
uacce->ops = NULL;
uacce->parent = NULL;
mutex_unlock(&uacce->mutex);
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
put_device(&uacce->dev);
}
EXPORT_SYMBOL_GPL(uacce_remove);
static int __init uacce_init(void)
{
int ret;
ret = class_register(&uacce_class);
if (ret)
return ret;
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
ret = alloc_chrdev_region(&uacce_devt, 0, MINORMASK, UACCE_NAME);
if (ret)
class_unregister(&uacce_class);
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
return ret;
}
static __exit void uacce_exit(void)
{
unregister_chrdev_region(uacce_devt, MINORMASK);
class_unregister(&uacce_class);
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
}
subsys_initcall(uacce_init);
module_exit(uacce_exit);
MODULE_LICENSE("GPL");
MODULE_AUTHOR("HiSilicon Tech. Co., Ltd.");
uacce: add uacce driver Uacce (Unified/User-space-access-intended Accelerator Framework) targets to provide Shared Virtual Addressing (SVA) between accelerators and processes. So accelerator can access any data structure of the main cpu. This differs from the data sharing between cpu and io device, which share only data content rather than address. Since unified address, hardware and user space of process can share the same virtual address in the communication. Uacce create a chrdev for every registration, the queue is allocated to the process when the chrdev is opened. Then the process can access the hardware resource by interact with the queue file. By mmap the queue file space to user space, the process can directly put requests to the hardware without syscall to the kernel space. The IOMMU core only tracks mm<->device bonds at the moment, because it only needs to handle IOTLB invalidation and PASID table entries. However uacce needs a finer granularity since multiple queues from the same device can be bound to an mm. When the mm exits, all bound queues must be stopped so that the IOMMU can safely clear the PASID table entry and reallocate the PASID. An intermediate struct uacce_mm links uacce devices and queues. Note that an mm may be bound to multiple devices but an uacce_mm structure only ever belongs to a single device, because we don't need anything more complex (if multiple devices are bound to one mm, then we'll create one uacce_mm for each bond). uacce_device --+-- uacce_mm --+-- uacce_queue | '-- uacce_queue | '-- uacce_mm --+-- uacce_queue +-- uacce_queue '-- uacce_queue Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Kenneth Lee <liguozhu@hisilicon.com> Signed-off-by: Zaibo Xu <xuzaibo@huawei.com> Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2020-02-11 07:54:23 +00:00
MODULE_DESCRIPTION("Accelerator interface for Userland applications");