linux/drivers/base/regmap/regcache-maple.c

402 lines
7.9 KiB
C
Raw Permalink Normal View History

regmap: Add maple tree based register cache The current state of the art for sparse register maps is the rbtree cache. This works well for most applications but isn't always ideal for sparser register maps since the rbtree can get deep, requiring a lot of walking. Fortunately the kernel has a data structure intended to address this very problem, the maple tree. Provide an initial implementation of a register cache based on the maple tree to start taking advantage of it. The entries stored in the maple tree are arrays of register values, with the maple tree keys holding the register addresses. We store data in host native format rather than device native format as we do for rbtree, this will be a benefit for devices where we don't marshal data within regmap and simplifies the code but will result in additional CPU overhead when syncing the cache on devices where we do marshal data in regmap. This should work well for a lot of devices, though there's some additional areas that could be looked at such as caching the last accessed entry like we do for rbtree and trying to minimise the maple tree level locking. We should also use bulk writes rather than single register writes when resyncing the cache where possible, even if we don't store in device native format. Very small register maps may continue to to better with rbtree longer term. Signed-off-by: Mark Brown <broonie@kernel.org> Link: https://lore.kernel.org/r/20230325-regcache-maple-v3-2-23e271f93dc7@kernel.org Signed-off-by: Mark Brown <broonie@kernel.org>
2023-03-30 00:10:24 +00:00
// SPDX-License-Identifier: GPL-2.0
//
// Register cache access API - maple tree based cache
//
// Copyright 2023 Arm, Ltd
//
// Author: Mark Brown <broonie@kernel.org>
#include <linux/debugfs.h>
#include <linux/device.h>
#include <linux/maple_tree.h>
#include <linux/slab.h>
#include "internal.h"
static int regcache_maple_read(struct regmap *map,
unsigned int reg, unsigned int *value)
{
struct maple_tree *mt = map->cache;
MA_STATE(mas, mt, reg, reg);
unsigned long *entry;
rcu_read_lock();
entry = mas_walk(&mas);
regmap: Add maple tree based register cache The current state of the art for sparse register maps is the rbtree cache. This works well for most applications but isn't always ideal for sparser register maps since the rbtree can get deep, requiring a lot of walking. Fortunately the kernel has a data structure intended to address this very problem, the maple tree. Provide an initial implementation of a register cache based on the maple tree to start taking advantage of it. The entries stored in the maple tree are arrays of register values, with the maple tree keys holding the register addresses. We store data in host native format rather than device native format as we do for rbtree, this will be a benefit for devices where we don't marshal data within regmap and simplifies the code but will result in additional CPU overhead when syncing the cache on devices where we do marshal data in regmap. This should work well for a lot of devices, though there's some additional areas that could be looked at such as caching the last accessed entry like we do for rbtree and trying to minimise the maple tree level locking. We should also use bulk writes rather than single register writes when resyncing the cache where possible, even if we don't store in device native format. Very small register maps may continue to to better with rbtree longer term. Signed-off-by: Mark Brown <broonie@kernel.org> Link: https://lore.kernel.org/r/20230325-regcache-maple-v3-2-23e271f93dc7@kernel.org Signed-off-by: Mark Brown <broonie@kernel.org>
2023-03-30 00:10:24 +00:00
if (!entry) {
rcu_read_unlock();
return -ENOENT;
}
*value = entry[reg - mas.index];
rcu_read_unlock();
return 0;
}
static int regcache_maple_write(struct regmap *map, unsigned int reg,
unsigned int val)
{
struct maple_tree *mt = map->cache;
MA_STATE(mas, mt, reg, reg);
unsigned long *entry, *upper, *lower;
unsigned long index, last;
size_t lower_sz, upper_sz;
int ret;
rcu_read_lock();
entry = mas_walk(&mas);
regmap: Add maple tree based register cache The current state of the art for sparse register maps is the rbtree cache. This works well for most applications but isn't always ideal for sparser register maps since the rbtree can get deep, requiring a lot of walking. Fortunately the kernel has a data structure intended to address this very problem, the maple tree. Provide an initial implementation of a register cache based on the maple tree to start taking advantage of it. The entries stored in the maple tree are arrays of register values, with the maple tree keys holding the register addresses. We store data in host native format rather than device native format as we do for rbtree, this will be a benefit for devices where we don't marshal data within regmap and simplifies the code but will result in additional CPU overhead when syncing the cache on devices where we do marshal data in regmap. This should work well for a lot of devices, though there's some additional areas that could be looked at such as caching the last accessed entry like we do for rbtree and trying to minimise the maple tree level locking. We should also use bulk writes rather than single register writes when resyncing the cache where possible, even if we don't store in device native format. Very small register maps may continue to to better with rbtree longer term. Signed-off-by: Mark Brown <broonie@kernel.org> Link: https://lore.kernel.org/r/20230325-regcache-maple-v3-2-23e271f93dc7@kernel.org Signed-off-by: Mark Brown <broonie@kernel.org>
2023-03-30 00:10:24 +00:00
if (entry) {
entry[reg - mas.index] = val;
rcu_read_unlock();
return 0;
}
/* Any adjacent entries to extend/merge? */
mas_set_range(&mas, reg - 1, reg + 1);
index = reg;
last = reg;
lower = mas_find(&mas, reg - 1);
if (lower) {
index = mas.index;
lower_sz = (mas.last - mas.index + 1) * sizeof(unsigned long);
}
upper = mas_find(&mas, reg + 1);
if (upper) {
last = mas.last;
upper_sz = (mas.last - mas.index + 1) * sizeof(unsigned long);
}
rcu_read_unlock();
entry = kmalloc((last - index + 1) * sizeof(unsigned long),
regmap: maple: Use alloc_flags for memory allocations REGCACHE_MAPLE needs to allocate memory for regmap operations. This results in lockdep splats if used with fast_io since fast_io uses spinlocks for locking. BUG: sleeping function called from invalid context at include/linux/sched/mm.h:306 in_atomic(): 1, irqs_disabled(): 128, non_block: 0, pid: 167, name: kunit_try_catch preempt_count: 1, expected: 0 1 lock held by kunit_try_catch/167: #0: 838e9c10 (regmap_kunit:86:(config)->lock){....}-{2:2}, at: regmap_lock_spinlock+0x14/0x1c irq event stamp: 146 hardirqs last enabled at (145): [<8078bfa8>] crng_make_state+0x1a0/0x294 hardirqs last disabled at (146): [<80c5f62c>] _raw_spin_lock_irqsave+0x7c/0x80 softirqs last enabled at (0): [<80110cc4>] copy_process+0x810/0x216c softirqs last disabled at (0): [<00000000>] 0x0 CPU: 0 PID: 167 Comm: kunit_try_catch Tainted: G N 6.5.0-rc1-00028-gc4be22597a36-dirty #6 Hardware name: Generic DT based system unwind_backtrace from show_stack+0x18/0x1c show_stack from dump_stack_lvl+0x38/0x5c dump_stack_lvl from __might_resched+0x188/0x2d0 __might_resched from __kmem_cache_alloc_node+0x1f4/0x258 __kmem_cache_alloc_node from __kmalloc+0x48/0x170 __kmalloc from regcache_maple_write+0x194/0x248 regcache_maple_write from _regmap_write+0x88/0x140 _regmap_write from regmap_write+0x44/0x68 regmap_write from basic_read_write+0x8c/0x27c basic_read_write from kunit_generic_run_threadfn_adapter+0x1c/0x28 kunit_generic_run_threadfn_adapter from kthread+0xf8/0x120 kthread from ret_from_fork+0x14/0x3c Exception stack(0x881a5fb0 to 0x881a5ff8) 5fa0: 00000000 00000000 00000000 00000000 5fc0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 5fe0: 00000000 00000000 00000000 00000000 00000013 00000000 Use map->alloc_flags instead of GFP_KERNEL for memory allocations to fix the problem. Fixes: f033c26de5a5 ("regmap: Add maple tree based register cache") Cc: Dan Carpenter <dan.carpenter@linaro.org> Signed-off-by: Guenter Roeck <linux@roeck-us.net> Link: https://lore.kernel.org/r/20230720172021.2617326-1-linux@roeck-us.net Signed-off-by: Mark Brown <broonie@kernel.org>
2023-07-20 17:20:21 +00:00
map->alloc_flags);
regmap: Add maple tree based register cache The current state of the art for sparse register maps is the rbtree cache. This works well for most applications but isn't always ideal for sparser register maps since the rbtree can get deep, requiring a lot of walking. Fortunately the kernel has a data structure intended to address this very problem, the maple tree. Provide an initial implementation of a register cache based on the maple tree to start taking advantage of it. The entries stored in the maple tree are arrays of register values, with the maple tree keys holding the register addresses. We store data in host native format rather than device native format as we do for rbtree, this will be a benefit for devices where we don't marshal data within regmap and simplifies the code but will result in additional CPU overhead when syncing the cache on devices where we do marshal data in regmap. This should work well for a lot of devices, though there's some additional areas that could be looked at such as caching the last accessed entry like we do for rbtree and trying to minimise the maple tree level locking. We should also use bulk writes rather than single register writes when resyncing the cache where possible, even if we don't store in device native format. Very small register maps may continue to to better with rbtree longer term. Signed-off-by: Mark Brown <broonie@kernel.org> Link: https://lore.kernel.org/r/20230325-regcache-maple-v3-2-23e271f93dc7@kernel.org Signed-off-by: Mark Brown <broonie@kernel.org>
2023-03-30 00:10:24 +00:00
if (!entry)
return -ENOMEM;
if (lower)
memcpy(entry, lower, lower_sz);
entry[reg - index] = val;
if (upper)
memcpy(&entry[reg - index + 1], upper, upper_sz);
/*
* This is safe because the regmap lock means the Maple lock
* is redundant, but we need to take it due to lockdep asserts
* in the maple tree code.
*/
mas_lock(&mas);
mas_set_range(&mas, index, last);
regmap: maple: Use alloc_flags for memory allocations REGCACHE_MAPLE needs to allocate memory for regmap operations. This results in lockdep splats if used with fast_io since fast_io uses spinlocks for locking. BUG: sleeping function called from invalid context at include/linux/sched/mm.h:306 in_atomic(): 1, irqs_disabled(): 128, non_block: 0, pid: 167, name: kunit_try_catch preempt_count: 1, expected: 0 1 lock held by kunit_try_catch/167: #0: 838e9c10 (regmap_kunit:86:(config)->lock){....}-{2:2}, at: regmap_lock_spinlock+0x14/0x1c irq event stamp: 146 hardirqs last enabled at (145): [<8078bfa8>] crng_make_state+0x1a0/0x294 hardirqs last disabled at (146): [<80c5f62c>] _raw_spin_lock_irqsave+0x7c/0x80 softirqs last enabled at (0): [<80110cc4>] copy_process+0x810/0x216c softirqs last disabled at (0): [<00000000>] 0x0 CPU: 0 PID: 167 Comm: kunit_try_catch Tainted: G N 6.5.0-rc1-00028-gc4be22597a36-dirty #6 Hardware name: Generic DT based system unwind_backtrace from show_stack+0x18/0x1c show_stack from dump_stack_lvl+0x38/0x5c dump_stack_lvl from __might_resched+0x188/0x2d0 __might_resched from __kmem_cache_alloc_node+0x1f4/0x258 __kmem_cache_alloc_node from __kmalloc+0x48/0x170 __kmalloc from regcache_maple_write+0x194/0x248 regcache_maple_write from _regmap_write+0x88/0x140 _regmap_write from regmap_write+0x44/0x68 regmap_write from basic_read_write+0x8c/0x27c basic_read_write from kunit_generic_run_threadfn_adapter+0x1c/0x28 kunit_generic_run_threadfn_adapter from kthread+0xf8/0x120 kthread from ret_from_fork+0x14/0x3c Exception stack(0x881a5fb0 to 0x881a5ff8) 5fa0: 00000000 00000000 00000000 00000000 5fc0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 5fe0: 00000000 00000000 00000000 00000000 00000013 00000000 Use map->alloc_flags instead of GFP_KERNEL for memory allocations to fix the problem. Fixes: f033c26de5a5 ("regmap: Add maple tree based register cache") Cc: Dan Carpenter <dan.carpenter@linaro.org> Signed-off-by: Guenter Roeck <linux@roeck-us.net> Link: https://lore.kernel.org/r/20230720172021.2617326-1-linux@roeck-us.net Signed-off-by: Mark Brown <broonie@kernel.org>
2023-07-20 17:20:21 +00:00
ret = mas_store_gfp(&mas, entry, map->alloc_flags);
regmap: Add maple tree based register cache The current state of the art for sparse register maps is the rbtree cache. This works well for most applications but isn't always ideal for sparser register maps since the rbtree can get deep, requiring a lot of walking. Fortunately the kernel has a data structure intended to address this very problem, the maple tree. Provide an initial implementation of a register cache based on the maple tree to start taking advantage of it. The entries stored in the maple tree are arrays of register values, with the maple tree keys holding the register addresses. We store data in host native format rather than device native format as we do for rbtree, this will be a benefit for devices where we don't marshal data within regmap and simplifies the code but will result in additional CPU overhead when syncing the cache on devices where we do marshal data in regmap. This should work well for a lot of devices, though there's some additional areas that could be looked at such as caching the last accessed entry like we do for rbtree and trying to minimise the maple tree level locking. We should also use bulk writes rather than single register writes when resyncing the cache where possible, even if we don't store in device native format. Very small register maps may continue to to better with rbtree longer term. Signed-off-by: Mark Brown <broonie@kernel.org> Link: https://lore.kernel.org/r/20230325-regcache-maple-v3-2-23e271f93dc7@kernel.org Signed-off-by: Mark Brown <broonie@kernel.org>
2023-03-30 00:10:24 +00:00
mas_unlock(&mas);
if (ret == 0) {
kfree(lower);
kfree(upper);
}
return ret;
}
static int regcache_maple_drop(struct regmap *map, unsigned int min,
unsigned int max)
{
struct maple_tree *mt = map->cache;
MA_STATE(mas, mt, min, max);
unsigned long *entry, *lower, *upper;
/* initialized to work around false-positive -Wuninitialized warning */
unsigned long lower_index = 0, lower_last = 0;
regmap: Add maple tree based register cache The current state of the art for sparse register maps is the rbtree cache. This works well for most applications but isn't always ideal for sparser register maps since the rbtree can get deep, requiring a lot of walking. Fortunately the kernel has a data structure intended to address this very problem, the maple tree. Provide an initial implementation of a register cache based on the maple tree to start taking advantage of it. The entries stored in the maple tree are arrays of register values, with the maple tree keys holding the register addresses. We store data in host native format rather than device native format as we do for rbtree, this will be a benefit for devices where we don't marshal data within regmap and simplifies the code but will result in additional CPU overhead when syncing the cache on devices where we do marshal data in regmap. This should work well for a lot of devices, though there's some additional areas that could be looked at such as caching the last accessed entry like we do for rbtree and trying to minimise the maple tree level locking. We should also use bulk writes rather than single register writes when resyncing the cache where possible, even if we don't store in device native format. Very small register maps may continue to to better with rbtree longer term. Signed-off-by: Mark Brown <broonie@kernel.org> Link: https://lore.kernel.org/r/20230325-regcache-maple-v3-2-23e271f93dc7@kernel.org Signed-off-by: Mark Brown <broonie@kernel.org>
2023-03-30 00:10:24 +00:00
unsigned long upper_index, upper_last;
int ret = 0;
regmap: Add maple tree based register cache The current state of the art for sparse register maps is the rbtree cache. This works well for most applications but isn't always ideal for sparser register maps since the rbtree can get deep, requiring a lot of walking. Fortunately the kernel has a data structure intended to address this very problem, the maple tree. Provide an initial implementation of a register cache based on the maple tree to start taking advantage of it. The entries stored in the maple tree are arrays of register values, with the maple tree keys holding the register addresses. We store data in host native format rather than device native format as we do for rbtree, this will be a benefit for devices where we don't marshal data within regmap and simplifies the code but will result in additional CPU overhead when syncing the cache on devices where we do marshal data in regmap. This should work well for a lot of devices, though there's some additional areas that could be looked at such as caching the last accessed entry like we do for rbtree and trying to minimise the maple tree level locking. We should also use bulk writes rather than single register writes when resyncing the cache where possible, even if we don't store in device native format. Very small register maps may continue to to better with rbtree longer term. Signed-off-by: Mark Brown <broonie@kernel.org> Link: https://lore.kernel.org/r/20230325-regcache-maple-v3-2-23e271f93dc7@kernel.org Signed-off-by: Mark Brown <broonie@kernel.org>
2023-03-30 00:10:24 +00:00
lower = NULL;
upper = NULL;
mas_lock(&mas);
mas_for_each(&mas, entry, max) {
/*
* This is safe because the regmap lock means the
* Maple lock is redundant, but we need to take it due
* to lockdep asserts in the maple tree code.
*/
mas_unlock(&mas);
/* Do we need to save any of this entry? */
if (mas.index < min) {
lower_index = mas.index;
lower_last = min -1;
lower = kmemdup_array(entry,
min - mas.index, sizeof(*lower),
map->alloc_flags);
regmap: Add maple tree based register cache The current state of the art for sparse register maps is the rbtree cache. This works well for most applications but isn't always ideal for sparser register maps since the rbtree can get deep, requiring a lot of walking. Fortunately the kernel has a data structure intended to address this very problem, the maple tree. Provide an initial implementation of a register cache based on the maple tree to start taking advantage of it. The entries stored in the maple tree are arrays of register values, with the maple tree keys holding the register addresses. We store data in host native format rather than device native format as we do for rbtree, this will be a benefit for devices where we don't marshal data within regmap and simplifies the code but will result in additional CPU overhead when syncing the cache on devices where we do marshal data in regmap. This should work well for a lot of devices, though there's some additional areas that could be looked at such as caching the last accessed entry like we do for rbtree and trying to minimise the maple tree level locking. We should also use bulk writes rather than single register writes when resyncing the cache where possible, even if we don't store in device native format. Very small register maps may continue to to better with rbtree longer term. Signed-off-by: Mark Brown <broonie@kernel.org> Link: https://lore.kernel.org/r/20230325-regcache-maple-v3-2-23e271f93dc7@kernel.org Signed-off-by: Mark Brown <broonie@kernel.org>
2023-03-30 00:10:24 +00:00
if (!lower) {
ret = -ENOMEM;
goto out_unlocked;
regmap: Add maple tree based register cache The current state of the art for sparse register maps is the rbtree cache. This works well for most applications but isn't always ideal for sparser register maps since the rbtree can get deep, requiring a lot of walking. Fortunately the kernel has a data structure intended to address this very problem, the maple tree. Provide an initial implementation of a register cache based on the maple tree to start taking advantage of it. The entries stored in the maple tree are arrays of register values, with the maple tree keys holding the register addresses. We store data in host native format rather than device native format as we do for rbtree, this will be a benefit for devices where we don't marshal data within regmap and simplifies the code but will result in additional CPU overhead when syncing the cache on devices where we do marshal data in regmap. This should work well for a lot of devices, though there's some additional areas that could be looked at such as caching the last accessed entry like we do for rbtree and trying to minimise the maple tree level locking. We should also use bulk writes rather than single register writes when resyncing the cache where possible, even if we don't store in device native format. Very small register maps may continue to to better with rbtree longer term. Signed-off-by: Mark Brown <broonie@kernel.org> Link: https://lore.kernel.org/r/20230325-regcache-maple-v3-2-23e271f93dc7@kernel.org Signed-off-by: Mark Brown <broonie@kernel.org>
2023-03-30 00:10:24 +00:00
}
}
if (mas.last > max) {
upper_index = max + 1;
upper_last = mas.last;
upper = kmemdup_array(&entry[max - mas.index + 1],
mas.last - max, sizeof(*upper),
map->alloc_flags);
regmap: Add maple tree based register cache The current state of the art for sparse register maps is the rbtree cache. This works well for most applications but isn't always ideal for sparser register maps since the rbtree can get deep, requiring a lot of walking. Fortunately the kernel has a data structure intended to address this very problem, the maple tree. Provide an initial implementation of a register cache based on the maple tree to start taking advantage of it. The entries stored in the maple tree are arrays of register values, with the maple tree keys holding the register addresses. We store data in host native format rather than device native format as we do for rbtree, this will be a benefit for devices where we don't marshal data within regmap and simplifies the code but will result in additional CPU overhead when syncing the cache on devices where we do marshal data in regmap. This should work well for a lot of devices, though there's some additional areas that could be looked at such as caching the last accessed entry like we do for rbtree and trying to minimise the maple tree level locking. We should also use bulk writes rather than single register writes when resyncing the cache where possible, even if we don't store in device native format. Very small register maps may continue to to better with rbtree longer term. Signed-off-by: Mark Brown <broonie@kernel.org> Link: https://lore.kernel.org/r/20230325-regcache-maple-v3-2-23e271f93dc7@kernel.org Signed-off-by: Mark Brown <broonie@kernel.org>
2023-03-30 00:10:24 +00:00
if (!upper) {
ret = -ENOMEM;
goto out_unlocked;
regmap: Add maple tree based register cache The current state of the art for sparse register maps is the rbtree cache. This works well for most applications but isn't always ideal for sparser register maps since the rbtree can get deep, requiring a lot of walking. Fortunately the kernel has a data structure intended to address this very problem, the maple tree. Provide an initial implementation of a register cache based on the maple tree to start taking advantage of it. The entries stored in the maple tree are arrays of register values, with the maple tree keys holding the register addresses. We store data in host native format rather than device native format as we do for rbtree, this will be a benefit for devices where we don't marshal data within regmap and simplifies the code but will result in additional CPU overhead when syncing the cache on devices where we do marshal data in regmap. This should work well for a lot of devices, though there's some additional areas that could be looked at such as caching the last accessed entry like we do for rbtree and trying to minimise the maple tree level locking. We should also use bulk writes rather than single register writes when resyncing the cache where possible, even if we don't store in device native format. Very small register maps may continue to to better with rbtree longer term. Signed-off-by: Mark Brown <broonie@kernel.org> Link: https://lore.kernel.org/r/20230325-regcache-maple-v3-2-23e271f93dc7@kernel.org Signed-off-by: Mark Brown <broonie@kernel.org>
2023-03-30 00:10:24 +00:00
}
}
kfree(entry);
mas_lock(&mas);
mas_erase(&mas);
/* Insert new nodes with the saved data */
if (lower) {
mas_set_range(&mas, lower_index, lower_last);
regmap: maple: Use alloc_flags for memory allocations REGCACHE_MAPLE needs to allocate memory for regmap operations. This results in lockdep splats if used with fast_io since fast_io uses spinlocks for locking. BUG: sleeping function called from invalid context at include/linux/sched/mm.h:306 in_atomic(): 1, irqs_disabled(): 128, non_block: 0, pid: 167, name: kunit_try_catch preempt_count: 1, expected: 0 1 lock held by kunit_try_catch/167: #0: 838e9c10 (regmap_kunit:86:(config)->lock){....}-{2:2}, at: regmap_lock_spinlock+0x14/0x1c irq event stamp: 146 hardirqs last enabled at (145): [<8078bfa8>] crng_make_state+0x1a0/0x294 hardirqs last disabled at (146): [<80c5f62c>] _raw_spin_lock_irqsave+0x7c/0x80 softirqs last enabled at (0): [<80110cc4>] copy_process+0x810/0x216c softirqs last disabled at (0): [<00000000>] 0x0 CPU: 0 PID: 167 Comm: kunit_try_catch Tainted: G N 6.5.0-rc1-00028-gc4be22597a36-dirty #6 Hardware name: Generic DT based system unwind_backtrace from show_stack+0x18/0x1c show_stack from dump_stack_lvl+0x38/0x5c dump_stack_lvl from __might_resched+0x188/0x2d0 __might_resched from __kmem_cache_alloc_node+0x1f4/0x258 __kmem_cache_alloc_node from __kmalloc+0x48/0x170 __kmalloc from regcache_maple_write+0x194/0x248 regcache_maple_write from _regmap_write+0x88/0x140 _regmap_write from regmap_write+0x44/0x68 regmap_write from basic_read_write+0x8c/0x27c basic_read_write from kunit_generic_run_threadfn_adapter+0x1c/0x28 kunit_generic_run_threadfn_adapter from kthread+0xf8/0x120 kthread from ret_from_fork+0x14/0x3c Exception stack(0x881a5fb0 to 0x881a5ff8) 5fa0: 00000000 00000000 00000000 00000000 5fc0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 5fe0: 00000000 00000000 00000000 00000000 00000013 00000000 Use map->alloc_flags instead of GFP_KERNEL for memory allocations to fix the problem. Fixes: f033c26de5a5 ("regmap: Add maple tree based register cache") Cc: Dan Carpenter <dan.carpenter@linaro.org> Signed-off-by: Guenter Roeck <linux@roeck-us.net> Link: https://lore.kernel.org/r/20230720172021.2617326-1-linux@roeck-us.net Signed-off-by: Mark Brown <broonie@kernel.org>
2023-07-20 17:20:21 +00:00
ret = mas_store_gfp(&mas, lower, map->alloc_flags);
regmap: Add maple tree based register cache The current state of the art for sparse register maps is the rbtree cache. This works well for most applications but isn't always ideal for sparser register maps since the rbtree can get deep, requiring a lot of walking. Fortunately the kernel has a data structure intended to address this very problem, the maple tree. Provide an initial implementation of a register cache based on the maple tree to start taking advantage of it. The entries stored in the maple tree are arrays of register values, with the maple tree keys holding the register addresses. We store data in host native format rather than device native format as we do for rbtree, this will be a benefit for devices where we don't marshal data within regmap and simplifies the code but will result in additional CPU overhead when syncing the cache on devices where we do marshal data in regmap. This should work well for a lot of devices, though there's some additional areas that could be looked at such as caching the last accessed entry like we do for rbtree and trying to minimise the maple tree level locking. We should also use bulk writes rather than single register writes when resyncing the cache where possible, even if we don't store in device native format. Very small register maps may continue to to better with rbtree longer term. Signed-off-by: Mark Brown <broonie@kernel.org> Link: https://lore.kernel.org/r/20230325-regcache-maple-v3-2-23e271f93dc7@kernel.org Signed-off-by: Mark Brown <broonie@kernel.org>
2023-03-30 00:10:24 +00:00
if (ret != 0)
goto out;
lower = NULL;
}
if (upper) {
mas_set_range(&mas, upper_index, upper_last);
regmap: maple: Use alloc_flags for memory allocations REGCACHE_MAPLE needs to allocate memory for regmap operations. This results in lockdep splats if used with fast_io since fast_io uses spinlocks for locking. BUG: sleeping function called from invalid context at include/linux/sched/mm.h:306 in_atomic(): 1, irqs_disabled(): 128, non_block: 0, pid: 167, name: kunit_try_catch preempt_count: 1, expected: 0 1 lock held by kunit_try_catch/167: #0: 838e9c10 (regmap_kunit:86:(config)->lock){....}-{2:2}, at: regmap_lock_spinlock+0x14/0x1c irq event stamp: 146 hardirqs last enabled at (145): [<8078bfa8>] crng_make_state+0x1a0/0x294 hardirqs last disabled at (146): [<80c5f62c>] _raw_spin_lock_irqsave+0x7c/0x80 softirqs last enabled at (0): [<80110cc4>] copy_process+0x810/0x216c softirqs last disabled at (0): [<00000000>] 0x0 CPU: 0 PID: 167 Comm: kunit_try_catch Tainted: G N 6.5.0-rc1-00028-gc4be22597a36-dirty #6 Hardware name: Generic DT based system unwind_backtrace from show_stack+0x18/0x1c show_stack from dump_stack_lvl+0x38/0x5c dump_stack_lvl from __might_resched+0x188/0x2d0 __might_resched from __kmem_cache_alloc_node+0x1f4/0x258 __kmem_cache_alloc_node from __kmalloc+0x48/0x170 __kmalloc from regcache_maple_write+0x194/0x248 regcache_maple_write from _regmap_write+0x88/0x140 _regmap_write from regmap_write+0x44/0x68 regmap_write from basic_read_write+0x8c/0x27c basic_read_write from kunit_generic_run_threadfn_adapter+0x1c/0x28 kunit_generic_run_threadfn_adapter from kthread+0xf8/0x120 kthread from ret_from_fork+0x14/0x3c Exception stack(0x881a5fb0 to 0x881a5ff8) 5fa0: 00000000 00000000 00000000 00000000 5fc0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 5fe0: 00000000 00000000 00000000 00000000 00000013 00000000 Use map->alloc_flags instead of GFP_KERNEL for memory allocations to fix the problem. Fixes: f033c26de5a5 ("regmap: Add maple tree based register cache") Cc: Dan Carpenter <dan.carpenter@linaro.org> Signed-off-by: Guenter Roeck <linux@roeck-us.net> Link: https://lore.kernel.org/r/20230720172021.2617326-1-linux@roeck-us.net Signed-off-by: Mark Brown <broonie@kernel.org>
2023-07-20 17:20:21 +00:00
ret = mas_store_gfp(&mas, upper, map->alloc_flags);
regmap: Add maple tree based register cache The current state of the art for sparse register maps is the rbtree cache. This works well for most applications but isn't always ideal for sparser register maps since the rbtree can get deep, requiring a lot of walking. Fortunately the kernel has a data structure intended to address this very problem, the maple tree. Provide an initial implementation of a register cache based on the maple tree to start taking advantage of it. The entries stored in the maple tree are arrays of register values, with the maple tree keys holding the register addresses. We store data in host native format rather than device native format as we do for rbtree, this will be a benefit for devices where we don't marshal data within regmap and simplifies the code but will result in additional CPU overhead when syncing the cache on devices where we do marshal data in regmap. This should work well for a lot of devices, though there's some additional areas that could be looked at such as caching the last accessed entry like we do for rbtree and trying to minimise the maple tree level locking. We should also use bulk writes rather than single register writes when resyncing the cache where possible, even if we don't store in device native format. Very small register maps may continue to to better with rbtree longer term. Signed-off-by: Mark Brown <broonie@kernel.org> Link: https://lore.kernel.org/r/20230325-regcache-maple-v3-2-23e271f93dc7@kernel.org Signed-off-by: Mark Brown <broonie@kernel.org>
2023-03-30 00:10:24 +00:00
if (ret != 0)
goto out;
upper = NULL;
}
}
out:
mas_unlock(&mas);
out_unlocked:
regmap: Add maple tree based register cache The current state of the art for sparse register maps is the rbtree cache. This works well for most applications but isn't always ideal for sparser register maps since the rbtree can get deep, requiring a lot of walking. Fortunately the kernel has a data structure intended to address this very problem, the maple tree. Provide an initial implementation of a register cache based on the maple tree to start taking advantage of it. The entries stored in the maple tree are arrays of register values, with the maple tree keys holding the register addresses. We store data in host native format rather than device native format as we do for rbtree, this will be a benefit for devices where we don't marshal data within regmap and simplifies the code but will result in additional CPU overhead when syncing the cache on devices where we do marshal data in regmap. This should work well for a lot of devices, though there's some additional areas that could be looked at such as caching the last accessed entry like we do for rbtree and trying to minimise the maple tree level locking. We should also use bulk writes rather than single register writes when resyncing the cache where possible, even if we don't store in device native format. Very small register maps may continue to to better with rbtree longer term. Signed-off-by: Mark Brown <broonie@kernel.org> Link: https://lore.kernel.org/r/20230325-regcache-maple-v3-2-23e271f93dc7@kernel.org Signed-off-by: Mark Brown <broonie@kernel.org>
2023-03-30 00:10:24 +00:00
kfree(lower);
kfree(upper);
return ret;
}
static int regcache_maple_sync_block(struct regmap *map, unsigned long *entry,
struct ma_state *mas,
unsigned int min, unsigned int max)
{
void *buf;
unsigned long r;
size_t val_bytes = map->format.val_bytes;
int ret = 0;
mas_pause(mas);
rcu_read_unlock();
/*
* Use a raw write if writing more than one register to a
* device that supports raw writes to reduce transaction
* overheads.
*/
if (max - min > 1 && regmap_can_raw_write(map)) {
buf = kmalloc(val_bytes * (max - min), map->alloc_flags);
if (!buf) {
ret = -ENOMEM;
goto out;
}
/* Render the data for a raw write */
for (r = min; r < max; r++) {
regcache_set_val(map, buf, r - min,
entry[r - mas->index]);
}
ret = _regmap_raw_write(map, min, buf, (max - min) * val_bytes,
false);
kfree(buf);
} else {
for (r = min; r < max; r++) {
ret = _regmap_write(map, r,
entry[r - mas->index]);
if (ret != 0)
goto out;
}
}
out:
rcu_read_lock();
return ret;
}
regmap: Add maple tree based register cache The current state of the art for sparse register maps is the rbtree cache. This works well for most applications but isn't always ideal for sparser register maps since the rbtree can get deep, requiring a lot of walking. Fortunately the kernel has a data structure intended to address this very problem, the maple tree. Provide an initial implementation of a register cache based on the maple tree to start taking advantage of it. The entries stored in the maple tree are arrays of register values, with the maple tree keys holding the register addresses. We store data in host native format rather than device native format as we do for rbtree, this will be a benefit for devices where we don't marshal data within regmap and simplifies the code but will result in additional CPU overhead when syncing the cache on devices where we do marshal data in regmap. This should work well for a lot of devices, though there's some additional areas that could be looked at such as caching the last accessed entry like we do for rbtree and trying to minimise the maple tree level locking. We should also use bulk writes rather than single register writes when resyncing the cache where possible, even if we don't store in device native format. Very small register maps may continue to to better with rbtree longer term. Signed-off-by: Mark Brown <broonie@kernel.org> Link: https://lore.kernel.org/r/20230325-regcache-maple-v3-2-23e271f93dc7@kernel.org Signed-off-by: Mark Brown <broonie@kernel.org>
2023-03-30 00:10:24 +00:00
static int regcache_maple_sync(struct regmap *map, unsigned int min,
unsigned int max)
{
struct maple_tree *mt = map->cache;
unsigned long *entry;
MA_STATE(mas, mt, min, max);
unsigned long lmin = min;
unsigned long lmax = max;
unsigned int r, v, sync_start;
int ret = 0;
bool sync_needed = false;
regmap: Add maple tree based register cache The current state of the art for sparse register maps is the rbtree cache. This works well for most applications but isn't always ideal for sparser register maps since the rbtree can get deep, requiring a lot of walking. Fortunately the kernel has a data structure intended to address this very problem, the maple tree. Provide an initial implementation of a register cache based on the maple tree to start taking advantage of it. The entries stored in the maple tree are arrays of register values, with the maple tree keys holding the register addresses. We store data in host native format rather than device native format as we do for rbtree, this will be a benefit for devices where we don't marshal data within regmap and simplifies the code but will result in additional CPU overhead when syncing the cache on devices where we do marshal data in regmap. This should work well for a lot of devices, though there's some additional areas that could be looked at such as caching the last accessed entry like we do for rbtree and trying to minimise the maple tree level locking. We should also use bulk writes rather than single register writes when resyncing the cache where possible, even if we don't store in device native format. Very small register maps may continue to to better with rbtree longer term. Signed-off-by: Mark Brown <broonie@kernel.org> Link: https://lore.kernel.org/r/20230325-regcache-maple-v3-2-23e271f93dc7@kernel.org Signed-off-by: Mark Brown <broonie@kernel.org>
2023-03-30 00:10:24 +00:00
map->cache_bypass = true;
rcu_read_lock();
mas_for_each(&mas, entry, max) {
for (r = max(mas.index, lmin); r <= min(mas.last, lmax); r++) {
v = entry[r - mas.index];
if (regcache_reg_needs_sync(map, r, v)) {
if (!sync_needed) {
sync_start = r;
sync_needed = true;
}
continue;
}
if (!sync_needed)
continue;
ret = regcache_maple_sync_block(map, entry, &mas,
sync_start, r);
regmap: Add maple tree based register cache The current state of the art for sparse register maps is the rbtree cache. This works well for most applications but isn't always ideal for sparser register maps since the rbtree can get deep, requiring a lot of walking. Fortunately the kernel has a data structure intended to address this very problem, the maple tree. Provide an initial implementation of a register cache based on the maple tree to start taking advantage of it. The entries stored in the maple tree are arrays of register values, with the maple tree keys holding the register addresses. We store data in host native format rather than device native format as we do for rbtree, this will be a benefit for devices where we don't marshal data within regmap and simplifies the code but will result in additional CPU overhead when syncing the cache on devices where we do marshal data in regmap. This should work well for a lot of devices, though there's some additional areas that could be looked at such as caching the last accessed entry like we do for rbtree and trying to minimise the maple tree level locking. We should also use bulk writes rather than single register writes when resyncing the cache where possible, even if we don't store in device native format. Very small register maps may continue to to better with rbtree longer term. Signed-off-by: Mark Brown <broonie@kernel.org> Link: https://lore.kernel.org/r/20230325-regcache-maple-v3-2-23e271f93dc7@kernel.org Signed-off-by: Mark Brown <broonie@kernel.org>
2023-03-30 00:10:24 +00:00
if (ret != 0)
goto out;
sync_needed = false;
}
if (sync_needed) {
ret = regcache_maple_sync_block(map, entry, &mas,
sync_start, r);
if (ret != 0)
goto out;
sync_needed = false;
regmap: Add maple tree based register cache The current state of the art for sparse register maps is the rbtree cache. This works well for most applications but isn't always ideal for sparser register maps since the rbtree can get deep, requiring a lot of walking. Fortunately the kernel has a data structure intended to address this very problem, the maple tree. Provide an initial implementation of a register cache based on the maple tree to start taking advantage of it. The entries stored in the maple tree are arrays of register values, with the maple tree keys holding the register addresses. We store data in host native format rather than device native format as we do for rbtree, this will be a benefit for devices where we don't marshal data within regmap and simplifies the code but will result in additional CPU overhead when syncing the cache on devices where we do marshal data in regmap. This should work well for a lot of devices, though there's some additional areas that could be looked at such as caching the last accessed entry like we do for rbtree and trying to minimise the maple tree level locking. We should also use bulk writes rather than single register writes when resyncing the cache where possible, even if we don't store in device native format. Very small register maps may continue to to better with rbtree longer term. Signed-off-by: Mark Brown <broonie@kernel.org> Link: https://lore.kernel.org/r/20230325-regcache-maple-v3-2-23e271f93dc7@kernel.org Signed-off-by: Mark Brown <broonie@kernel.org>
2023-03-30 00:10:24 +00:00
}
}
out:
regmap: Add maple tree based register cache The current state of the art for sparse register maps is the rbtree cache. This works well for most applications but isn't always ideal for sparser register maps since the rbtree can get deep, requiring a lot of walking. Fortunately the kernel has a data structure intended to address this very problem, the maple tree. Provide an initial implementation of a register cache based on the maple tree to start taking advantage of it. The entries stored in the maple tree are arrays of register values, with the maple tree keys holding the register addresses. We store data in host native format rather than device native format as we do for rbtree, this will be a benefit for devices where we don't marshal data within regmap and simplifies the code but will result in additional CPU overhead when syncing the cache on devices where we do marshal data in regmap. This should work well for a lot of devices, though there's some additional areas that could be looked at such as caching the last accessed entry like we do for rbtree and trying to minimise the maple tree level locking. We should also use bulk writes rather than single register writes when resyncing the cache where possible, even if we don't store in device native format. Very small register maps may continue to to better with rbtree longer term. Signed-off-by: Mark Brown <broonie@kernel.org> Link: https://lore.kernel.org/r/20230325-regcache-maple-v3-2-23e271f93dc7@kernel.org Signed-off-by: Mark Brown <broonie@kernel.org>
2023-03-30 00:10:24 +00:00
rcu_read_unlock();
map->cache_bypass = false;
return ret;
}
static int regcache_maple_exit(struct regmap *map)
{
struct maple_tree *mt = map->cache;
MA_STATE(mas, mt, 0, UINT_MAX);
unsigned int *entry;
regmap: Add maple tree based register cache The current state of the art for sparse register maps is the rbtree cache. This works well for most applications but isn't always ideal for sparser register maps since the rbtree can get deep, requiring a lot of walking. Fortunately the kernel has a data structure intended to address this very problem, the maple tree. Provide an initial implementation of a register cache based on the maple tree to start taking advantage of it. The entries stored in the maple tree are arrays of register values, with the maple tree keys holding the register addresses. We store data in host native format rather than device native format as we do for rbtree, this will be a benefit for devices where we don't marshal data within regmap and simplifies the code but will result in additional CPU overhead when syncing the cache on devices where we do marshal data in regmap. This should work well for a lot of devices, though there's some additional areas that could be looked at such as caching the last accessed entry like we do for rbtree and trying to minimise the maple tree level locking. We should also use bulk writes rather than single register writes when resyncing the cache where possible, even if we don't store in device native format. Very small register maps may continue to to better with rbtree longer term. Signed-off-by: Mark Brown <broonie@kernel.org> Link: https://lore.kernel.org/r/20230325-regcache-maple-v3-2-23e271f93dc7@kernel.org Signed-off-by: Mark Brown <broonie@kernel.org>
2023-03-30 00:10:24 +00:00
/* if we've already been called then just return */
if (!mt)
return 0;
mas_lock(&mas);
mas_for_each(&mas, entry, UINT_MAX)
kfree(entry);
__mt_destroy(mt);
mas_unlock(&mas);
kfree(mt);
map->cache = NULL;
return 0;
}
static int regcache_maple_insert_block(struct regmap *map, int first,
int last)
{
struct maple_tree *mt = map->cache;
MA_STATE(mas, mt, first, last);
unsigned long *entry;
int i, ret;
regmap: maple: Use alloc_flags for memory allocations REGCACHE_MAPLE needs to allocate memory for regmap operations. This results in lockdep splats if used with fast_io since fast_io uses spinlocks for locking. BUG: sleeping function called from invalid context at include/linux/sched/mm.h:306 in_atomic(): 1, irqs_disabled(): 128, non_block: 0, pid: 167, name: kunit_try_catch preempt_count: 1, expected: 0 1 lock held by kunit_try_catch/167: #0: 838e9c10 (regmap_kunit:86:(config)->lock){....}-{2:2}, at: regmap_lock_spinlock+0x14/0x1c irq event stamp: 146 hardirqs last enabled at (145): [<8078bfa8>] crng_make_state+0x1a0/0x294 hardirqs last disabled at (146): [<80c5f62c>] _raw_spin_lock_irqsave+0x7c/0x80 softirqs last enabled at (0): [<80110cc4>] copy_process+0x810/0x216c softirqs last disabled at (0): [<00000000>] 0x0 CPU: 0 PID: 167 Comm: kunit_try_catch Tainted: G N 6.5.0-rc1-00028-gc4be22597a36-dirty #6 Hardware name: Generic DT based system unwind_backtrace from show_stack+0x18/0x1c show_stack from dump_stack_lvl+0x38/0x5c dump_stack_lvl from __might_resched+0x188/0x2d0 __might_resched from __kmem_cache_alloc_node+0x1f4/0x258 __kmem_cache_alloc_node from __kmalloc+0x48/0x170 __kmalloc from regcache_maple_write+0x194/0x248 regcache_maple_write from _regmap_write+0x88/0x140 _regmap_write from regmap_write+0x44/0x68 regmap_write from basic_read_write+0x8c/0x27c basic_read_write from kunit_generic_run_threadfn_adapter+0x1c/0x28 kunit_generic_run_threadfn_adapter from kthread+0xf8/0x120 kthread from ret_from_fork+0x14/0x3c Exception stack(0x881a5fb0 to 0x881a5ff8) 5fa0: 00000000 00000000 00000000 00000000 5fc0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 5fe0: 00000000 00000000 00000000 00000000 00000013 00000000 Use map->alloc_flags instead of GFP_KERNEL for memory allocations to fix the problem. Fixes: f033c26de5a5 ("regmap: Add maple tree based register cache") Cc: Dan Carpenter <dan.carpenter@linaro.org> Signed-off-by: Guenter Roeck <linux@roeck-us.net> Link: https://lore.kernel.org/r/20230720172021.2617326-1-linux@roeck-us.net Signed-off-by: Mark Brown <broonie@kernel.org>
2023-07-20 17:20:21 +00:00
entry = kcalloc(last - first + 1, sizeof(unsigned long), map->alloc_flags);
if (!entry)
return -ENOMEM;
for (i = 0; i < last - first + 1; i++)
entry[i] = map->reg_defaults[first + i].def;
mas_lock(&mas);
mas_set_range(&mas, map->reg_defaults[first].reg,
map->reg_defaults[last].reg);
regmap: maple: Use alloc_flags for memory allocations REGCACHE_MAPLE needs to allocate memory for regmap operations. This results in lockdep splats if used with fast_io since fast_io uses spinlocks for locking. BUG: sleeping function called from invalid context at include/linux/sched/mm.h:306 in_atomic(): 1, irqs_disabled(): 128, non_block: 0, pid: 167, name: kunit_try_catch preempt_count: 1, expected: 0 1 lock held by kunit_try_catch/167: #0: 838e9c10 (regmap_kunit:86:(config)->lock){....}-{2:2}, at: regmap_lock_spinlock+0x14/0x1c irq event stamp: 146 hardirqs last enabled at (145): [<8078bfa8>] crng_make_state+0x1a0/0x294 hardirqs last disabled at (146): [<80c5f62c>] _raw_spin_lock_irqsave+0x7c/0x80 softirqs last enabled at (0): [<80110cc4>] copy_process+0x810/0x216c softirqs last disabled at (0): [<00000000>] 0x0 CPU: 0 PID: 167 Comm: kunit_try_catch Tainted: G N 6.5.0-rc1-00028-gc4be22597a36-dirty #6 Hardware name: Generic DT based system unwind_backtrace from show_stack+0x18/0x1c show_stack from dump_stack_lvl+0x38/0x5c dump_stack_lvl from __might_resched+0x188/0x2d0 __might_resched from __kmem_cache_alloc_node+0x1f4/0x258 __kmem_cache_alloc_node from __kmalloc+0x48/0x170 __kmalloc from regcache_maple_write+0x194/0x248 regcache_maple_write from _regmap_write+0x88/0x140 _regmap_write from regmap_write+0x44/0x68 regmap_write from basic_read_write+0x8c/0x27c basic_read_write from kunit_generic_run_threadfn_adapter+0x1c/0x28 kunit_generic_run_threadfn_adapter from kthread+0xf8/0x120 kthread from ret_from_fork+0x14/0x3c Exception stack(0x881a5fb0 to 0x881a5ff8) 5fa0: 00000000 00000000 00000000 00000000 5fc0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 5fe0: 00000000 00000000 00000000 00000000 00000013 00000000 Use map->alloc_flags instead of GFP_KERNEL for memory allocations to fix the problem. Fixes: f033c26de5a5 ("regmap: Add maple tree based register cache") Cc: Dan Carpenter <dan.carpenter@linaro.org> Signed-off-by: Guenter Roeck <linux@roeck-us.net> Link: https://lore.kernel.org/r/20230720172021.2617326-1-linux@roeck-us.net Signed-off-by: Mark Brown <broonie@kernel.org>
2023-07-20 17:20:21 +00:00
ret = mas_store_gfp(&mas, entry, map->alloc_flags);
mas_unlock(&mas);
if (ret)
kfree(entry);
return ret;
}
regmap: Add maple tree based register cache The current state of the art for sparse register maps is the rbtree cache. This works well for most applications but isn't always ideal for sparser register maps since the rbtree can get deep, requiring a lot of walking. Fortunately the kernel has a data structure intended to address this very problem, the maple tree. Provide an initial implementation of a register cache based on the maple tree to start taking advantage of it. The entries stored in the maple tree are arrays of register values, with the maple tree keys holding the register addresses. We store data in host native format rather than device native format as we do for rbtree, this will be a benefit for devices where we don't marshal data within regmap and simplifies the code but will result in additional CPU overhead when syncing the cache on devices where we do marshal data in regmap. This should work well for a lot of devices, though there's some additional areas that could be looked at such as caching the last accessed entry like we do for rbtree and trying to minimise the maple tree level locking. We should also use bulk writes rather than single register writes when resyncing the cache where possible, even if we don't store in device native format. Very small register maps may continue to to better with rbtree longer term. Signed-off-by: Mark Brown <broonie@kernel.org> Link: https://lore.kernel.org/r/20230325-regcache-maple-v3-2-23e271f93dc7@kernel.org Signed-off-by: Mark Brown <broonie@kernel.org>
2023-03-30 00:10:24 +00:00
static int regcache_maple_init(struct regmap *map)
{
struct maple_tree *mt;
int i;
int ret;
int range_start;
regmap: Add maple tree based register cache The current state of the art for sparse register maps is the rbtree cache. This works well for most applications but isn't always ideal for sparser register maps since the rbtree can get deep, requiring a lot of walking. Fortunately the kernel has a data structure intended to address this very problem, the maple tree. Provide an initial implementation of a register cache based on the maple tree to start taking advantage of it. The entries stored in the maple tree are arrays of register values, with the maple tree keys holding the register addresses. We store data in host native format rather than device native format as we do for rbtree, this will be a benefit for devices where we don't marshal data within regmap and simplifies the code but will result in additional CPU overhead when syncing the cache on devices where we do marshal data in regmap. This should work well for a lot of devices, though there's some additional areas that could be looked at such as caching the last accessed entry like we do for rbtree and trying to minimise the maple tree level locking. We should also use bulk writes rather than single register writes when resyncing the cache where possible, even if we don't store in device native format. Very small register maps may continue to to better with rbtree longer term. Signed-off-by: Mark Brown <broonie@kernel.org> Link: https://lore.kernel.org/r/20230325-regcache-maple-v3-2-23e271f93dc7@kernel.org Signed-off-by: Mark Brown <broonie@kernel.org>
2023-03-30 00:10:24 +00:00
mt = kmalloc(sizeof(*mt), map->alloc_flags);
regmap: Add maple tree based register cache The current state of the art for sparse register maps is the rbtree cache. This works well for most applications but isn't always ideal for sparser register maps since the rbtree can get deep, requiring a lot of walking. Fortunately the kernel has a data structure intended to address this very problem, the maple tree. Provide an initial implementation of a register cache based on the maple tree to start taking advantage of it. The entries stored in the maple tree are arrays of register values, with the maple tree keys holding the register addresses. We store data in host native format rather than device native format as we do for rbtree, this will be a benefit for devices where we don't marshal data within regmap and simplifies the code but will result in additional CPU overhead when syncing the cache on devices where we do marshal data in regmap. This should work well for a lot of devices, though there's some additional areas that could be looked at such as caching the last accessed entry like we do for rbtree and trying to minimise the maple tree level locking. We should also use bulk writes rather than single register writes when resyncing the cache where possible, even if we don't store in device native format. Very small register maps may continue to to better with rbtree longer term. Signed-off-by: Mark Brown <broonie@kernel.org> Link: https://lore.kernel.org/r/20230325-regcache-maple-v3-2-23e271f93dc7@kernel.org Signed-off-by: Mark Brown <broonie@kernel.org>
2023-03-30 00:10:24 +00:00
if (!mt)
return -ENOMEM;
map->cache = mt;
mt_init(mt);
regmap: maple: Provide lockdep (sub)class for maple tree's internal lock In some cases when using the maple tree register cache, the lockdep validator might complain about invalid deadlocks: [7.131886] Possible interrupt unsafe locking scenario: [7.131890] CPU0 CPU1 [7.131893] ---- ---- [7.131896] lock(&mt->ma_lock); [7.131904] local_irq_disable(); [7.131907] lock(rockchip_drm_vop2:3114:(&vop2_regmap_config)->lock); [7.131916] lock(&mt->ma_lock); [7.131925] <Interrupt> [7.131928] lock(rockchip_drm_vop2:3114:(&vop2_regmap_config)->lock); [7.131936] *** DEADLOCK *** [7.131939] no locks held by swapper/0/0. [7.131944] the shortest dependencies between 2nd lock and 1st lock: [7.131950] -> (&mt->ma_lock){+.+.}-{2:2} { [7.131966] HARDIRQ-ON-W at: [7.131973] lock_acquire+0x200/0x330 [7.131986] _raw_spin_lock+0x50/0x70 [7.131998] regcache_maple_write+0x68/0xe0 [7.132010] regcache_write+0x6c/0x90 [7.132019] _regmap_read+0x19c/0x1d0 [7.132029] _regmap_update_bits+0xc0/0x148 [7.132038] regmap_update_bits_base+0x6c/0xa8 [7.132048] rk8xx_probe+0x22c/0x3d8 [7.132057] rk8xx_spi_probe+0x74/0x88 [7.132065] spi_probe+0xa8/0xe0 [...] [7.132675] } [7.132678] ... key at: [<ffff800082943c20>] __key.0+0x0/0x10 [7.132691] ... acquired at: [7.132695] _raw_spin_lock+0x50/0x70 [7.132704] regcache_maple_write+0x68/0xe0 [7.132714] regcache_write+0x6c/0x90 [7.132724] _regmap_read+0x19c/0x1d0 [7.132732] _regmap_update_bits+0xc0/0x148 [7.132741] regmap_field_update_bits_base+0x74/0xb8 [7.132751] vop2_plane_atomic_update+0x480/0x14d8 [rockchipdrm] [7.132820] drm_atomic_helper_commit_planes+0x1a0/0x320 [drm_kms_helper] [...] [7.135112] -> (rockchip_drm_vop2:3114:(&vop2_regmap_config)->lock){-...}-{2:2} { [7.135130] IN-HARDIRQ-W at: [7.135136] lock_acquire+0x200/0x330 [7.135147] _raw_spin_lock_irqsave+0x6c/0x98 [7.135157] regmap_lock_spinlock+0x20/0x40 [7.135166] regmap_read+0x44/0x90 [7.135175] vop2_isr+0x90/0x290 [rockchipdrm] [7.135225] __handle_irq_event_percpu+0x124/0x2d0 In the example above, the validator seems to get the scope of dependencies wrong, since the regmap instance used in rk8xx-spi driver has nothing to do with the instance from vop2. Improve validation by sharing the regmap's lockdep class with the maple tree's internal lock, while also providing a subclass for the latter. Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com> Link: https://patch.msgid.link/20241031-regmap-maple-lockdep-fix-v2-1-06a3710f3623@collabora.com Signed-off-by: Mark Brown <broonie@kernel.org>
2024-10-31 16:37:04 +00:00
if (!mt_external_lock(mt) && map->lock_key)
lockdep_set_class_and_subclass(&mt->ma_lock, map->lock_key, 1);
if (!map->num_reg_defaults)
return 0;
range_start = 0;
/* Scan for ranges of contiguous registers */
for (i = 1; i < map->num_reg_defaults; i++) {
if (map->reg_defaults[i].reg !=
map->reg_defaults[i - 1].reg + 1) {
ret = regcache_maple_insert_block(map, range_start,
i - 1);
if (ret != 0)
goto err;
range_start = i;
}
regmap: Add maple tree based register cache The current state of the art for sparse register maps is the rbtree cache. This works well for most applications but isn't always ideal for sparser register maps since the rbtree can get deep, requiring a lot of walking. Fortunately the kernel has a data structure intended to address this very problem, the maple tree. Provide an initial implementation of a register cache based on the maple tree to start taking advantage of it. The entries stored in the maple tree are arrays of register values, with the maple tree keys holding the register addresses. We store data in host native format rather than device native format as we do for rbtree, this will be a benefit for devices where we don't marshal data within regmap and simplifies the code but will result in additional CPU overhead when syncing the cache on devices where we do marshal data in regmap. This should work well for a lot of devices, though there's some additional areas that could be looked at such as caching the last accessed entry like we do for rbtree and trying to minimise the maple tree level locking. We should also use bulk writes rather than single register writes when resyncing the cache where possible, even if we don't store in device native format. Very small register maps may continue to to better with rbtree longer term. Signed-off-by: Mark Brown <broonie@kernel.org> Link: https://lore.kernel.org/r/20230325-regcache-maple-v3-2-23e271f93dc7@kernel.org Signed-off-by: Mark Brown <broonie@kernel.org>
2023-03-30 00:10:24 +00:00
}
/* Add the last block */
ret = regcache_maple_insert_block(map, range_start,
map->num_reg_defaults - 1);
if (ret != 0)
goto err;
regmap: Add maple tree based register cache The current state of the art for sparse register maps is the rbtree cache. This works well for most applications but isn't always ideal for sparser register maps since the rbtree can get deep, requiring a lot of walking. Fortunately the kernel has a data structure intended to address this very problem, the maple tree. Provide an initial implementation of a register cache based on the maple tree to start taking advantage of it. The entries stored in the maple tree are arrays of register values, with the maple tree keys holding the register addresses. We store data in host native format rather than device native format as we do for rbtree, this will be a benefit for devices where we don't marshal data within regmap and simplifies the code but will result in additional CPU overhead when syncing the cache on devices where we do marshal data in regmap. This should work well for a lot of devices, though there's some additional areas that could be looked at such as caching the last accessed entry like we do for rbtree and trying to minimise the maple tree level locking. We should also use bulk writes rather than single register writes when resyncing the cache where possible, even if we don't store in device native format. Very small register maps may continue to to better with rbtree longer term. Signed-off-by: Mark Brown <broonie@kernel.org> Link: https://lore.kernel.org/r/20230325-regcache-maple-v3-2-23e271f93dc7@kernel.org Signed-off-by: Mark Brown <broonie@kernel.org>
2023-03-30 00:10:24 +00:00
return 0;
err:
regcache_maple_exit(map);
return ret;
}
struct regcache_ops regcache_maple_ops = {
.type = REGCACHE_MAPLE,
.name = "maple",
.init = regcache_maple_init,
.exit = regcache_maple_exit,
.read = regcache_maple_read,
.write = regcache_maple_write,
.drop = regcache_maple_drop,
.sync = regcache_maple_sync,
};