lib/stackdepot: Double DEPOT_POOLS_CAP if KASAN is enabled
PATCH v2 1/1 lib/stackdepot: Double DEPOT_POOLS_CAP if KASAN is enabled
As said in commit 02754e0a484a ("lib/stackdepot.c: bump stackdepot capacity from 16MB to 128MB"), KASAN uses stackdepot to memorize stacks for all kmalloc/kfree calls. So stackdepot capacity was increased 8 times to accommodate KASAN usage even thought it was claimed 4X should be enough at that time.
With commit fc60e0caa94d ("lib/stackdepot: use fixed-sized slots for stack records"), all stackdepot records uses a fixed size with CONFIG_STACKDEPOT_MAX_FRAMES (default=64) entries. This is merged to support evictable KASAN stack records. Commit 31639fd6cebd ("stackdepot:
use variable size records for non-evictable entries") re-enabled the use of variable size records for non-KASAN use cases, but KASAN (generic mode) still uses the large fixed size stack records.
With the default CONFIG_STACKDEPOT_MAX_FRAMES of 64, KASAN use of stackdepot space had been more than double than before. Assuming an average stack frame size of 16, a KASAN stack record is almost 4X the size of a non-KASAN one.
When a wide variety of workloads are run on a debug kernel with KASAN enabled, the following warning may sometimes be printed.
[ 6818.650674] Stack depot reached limit capacity [ 6818.650730] WARNING: CPU: 1 PID: 272741 at lib/stackdepot.c:252 depot_alloc_stack+0x39e/0x3d0 : [ 6818.650907] Call Trace: [ 6818.650909] [<00047dd453d84b92>] depot_alloc_stack+0x3a2/0x3d0 [ 6818.650916] [<00047dd453d85254>] stack_depot_save_flags+0x4f4/0x5c0 [ 6818.650920] [<00047dd4535872c6>] kasan_save_stack+0x56/0x70 [ 6818.650924] [<00047dd453587328>] kasan_save_track+0x28/0x40 [ 6818.650927] [<00047dd45358a27a>] kasan_save_free_info+0x4a/0x70 [ 6818.650930] [<00047dd45358766a>] __kasan_slab_free+0x12a/0x1d0 [ 6818.650933] [<00047dd45350deb4>] kmem_cache_free+0x1b4/0x580 [ 6818.650938] [<00047dd452c520da>] __put_task_struct+0x24a/0x320 [ 6818.650945] [<00047dd452c6aee4>] delayed_put_task_struct+0x294/0x350 [ 6818.650949] [<00047dd452e9066a>] rcu_do_batch+0x6ea/0x2090 [ 6818.650953] [<00047dd452ea60f4>] rcu_core+0x474/0xa90 [ 6818.650956] [<00047dd452c780c0>] handle_softirqs+0x3c0/0xf90 [ 6818.650960] [<00047dd452c76fbe>] __irq_exit_rcu+0x35e/0x460 [ 6818.650963] [<00047dd452c79992>] irq_exit_rcu+0x22/0xb0 [ 6818.650966] [<00047dd454bd8128>] do_ext_irq+0xd8/0x120 [ 6818.650972] [<00047dd454c0ddd0>] ext_int_handler+0xb8/0xe8 [ 6818.650979] [<00047dd453589cf6>] kasan_check_range+0x236/0x2f0 [ 6818.650982] [<00047dd453378cf0>] filemap_get_pages+0x190/0xaa0 [ 6818.650986] [<00047dd453379940>] filemap_read+0x340/0xa70 [ 6818.650989] [<00047dd3d325d226>] xfs_file_buffered_read+0x2c6/0x400 [xfs] [ 6818.651431] [<00047dd3d325dfe2>] xfs_file_read_iter+0x2c2/0x550 [xfs] [ 6818.651663] [<00047dd45364710c>] vfs_read+0x64c/0x8c0 [ 6818.651669] [<00047dd453648ed8>] ksys_read+0x118/0x200 [ 6818.651672] [<00047dd452b6cf5a>] do_syscall+0x27a/0x380 [ 6818.651676] [<00047dd454bd7e74>] __do_syscall+0xf4/0x1a0 [ 6818.651680] [<00047dd454c0db58>] system_call+0x70/0x98
With all the recent changes in stackdepot to support new KASAN features, it is obvious that the current DEPOT_POOLS_CAP of 8192 may not be enough when KASAN is enabled. Fix this stackdepot capability issue by doubling DEPOT_POOLS_CAP if KASAN is enabled. With 4k pages, the maximum stackdepot capacity is doubled to 256 MB with KASAN enabled.
Also use the MIN() macro for defining DEPOT_MAX_POOLS to clarify the intention.
--- lib/stackdepot.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-)
lib/stackdepot.c +4 -3
@@ -36,11 +36,12 @@#include <linux/memblock.h> #include <linux/kasan-enabled.h> -#define DEPOT_POOLS_CAP 8192 +/* KASAN is a big user of stackdepot, double the cap if KASAN is enabled */ +#define DEPOT_POOLS_CAP (8192 * (IS_ENABLED(CONFIG_KASAN) ? 2 : 1)) +/* The pool_index is offset by 1 so the first record does not have a 0 handle. */ #define DEPOT_MAX_POOLS \ - (((1LL << (DEPOT_POOL_INDEX_BITS)) - 1 < DEPOT_POOLS_CAP) ? \ - (1LL << (DEPOT_POOL_INDEX_BITS)) - 1 : DEPOT_POOLS_CAP) + MIN((1LL << (DEPOT_POOL_INDEX_BITS)) - 1, DEPOT_POOLS_CAP)static bool stack_depot_disabled;static bool __stack_depot_early_init_requested__initdata =IS_ENABLED (CONFIG_STACKDEPOT_ALWAYS_INIT );
Signature · 1 line
2.43.5
Quoted 78 lines · Waiman Long
As said in commit
02754e0a484a("lib/stackdepot.c: bump stackdepot capacity from 16MB to 128MB"), KASAN uses stackdepot to memorize stacks for all kmalloc/kfree calls. So stackdepot capacity was increased 8 times to accommodate KASAN usage even thought it was claimed 4X should be enough at that time.With commit
fc60e0caa94d("lib/stackdepot: use fixed-sized slots for stack records"), all stackdepot records uses a fixed size withCONFIG_STACKDEPOT_MAX_FRAMES(default=64) entries. This is merged to support evictable KASAN stack records. Commit31639fd6cebd("stackdepot:
use variable size records for non-evictable entries") re-enabled the use of variable size records for non-KASAN use cases, but KASAN (generic mode) still uses the large fixed size stack records.With the default
CONFIG_STACKDEPOT_MAX_FRAMESof 64, KASAN use of stackdepot space had been more than double than before. Assuming an average stack frame size of 16, a KASAN stack record is almost 4X the size of a non-KASAN one.When a wide variety of workloads are run on a debug kernel with KASAN enabled, the following warning may sometimes be printed.
[ 6818.650674] Stack depot reached limit capacity [ 6818.650730] WARNING: CPU: 1 PID: 272741 at lib/stackdepot.c:252 depot_alloc_stack+0x39e/0x3d0 : [ 6818.650907] Call Trace: [ 6818.650909] [<00047dd453d84b92>] depot_alloc_stack+0x3a2/0x3d0 [ 6818.650916] [<00047dd453d85254>] stack_depot_save_flags+0x4f4/0x5c0 [ 6818.650920] [<00047dd4535872c6>] kasan_save_stack+0x56/0x70 [ 6818.650924] [<00047dd453587328>] kasan_save_track+0x28/0x40 [ 6818.650927] [<00047dd45358a27a>] kasan_save_free_info+0x4a/0x70 [ 6818.650930] [<00047dd45358766a>] __kasan_slab_free+0x12a/0x1d0 [ 6818.650933] [<00047dd45350deb4>] kmem_cache_free+0x1b4/0x580 [ 6818.650938] [<00047dd452c520da>] __put_task_struct+0x24a/0x320 [ 6818.650945] [<00047dd452c6aee4>] delayed_put_task_struct+0x294/0x350 [ 6818.650949] [<00047dd452e9066a>] rcu_do_batch+0x6ea/0x2090 [ 6818.650953] [<00047dd452ea60f4>] rcu_core+0x474/0xa90 [ 6818.650956] [<00047dd452c780c0>] handle_softirqs+0x3c0/0xf90 [ 6818.650960] [<00047dd452c76fbe>] __irq_exit_rcu+0x35e/0x460 [ 6818.650963] [<00047dd452c79992>] irq_exit_rcu+0x22/0xb0 [ 6818.650966] [<00047dd454bd8128>] do_ext_irq+0xd8/0x120 [ 6818.650972] [<00047dd454c0ddd0>] ext_int_handler+0xb8/0xe8 [ 6818.650979] [<00047dd453589cf6>] kasan_check_range+0x236/0x2f0 [ 6818.650982] [<00047dd453378cf0>] filemap_get_pages+0x190/0xaa0 [ 6818.650986] [<00047dd453379940>] filemap_read+0x340/0xa70 [ 6818.650989] [<00047dd3d325d226>] xfs_file_buffered_read+0x2c6/0x400 [xfs] [ 6818.651431] [<00047dd3d325dfe2>] xfs_file_read_iter+0x2c2/0x550 [xfs] [ 6818.651663] [<00047dd45364710c>] vfs_read+0x64c/0x8c0 [ 6818.651669] [<00047dd453648ed8>] ksys_read+0x118/0x200 [ 6818.651672] [<00047dd452b6cf5a>] do_syscall+0x27a/0x380 [ 6818.651676] [<00047dd454bd7e74>] __do_syscall+0xf4/0x1a0 [ 6818.651680] [<00047dd454c0db58>] system_call+0x70/0x98With all the recent changes in stackdepot to support new KASAN features, it is obvious that the current
DEPOT_POOLS_CAPof 8192 may not be enough when KASAN is enabled. Fix this stackdepot capability issue by doublingDEPOT_POOLS_CAPif KASAN is enabled. With 4k pages, the maximum stackdepot capacity is doubled to 256 MB with KASAN enabled.Also use the
MIN()macro for definingDEPOT_MAX_POOLSto clarify the intention.Fixes fc60e0caa94d ("lib/stackdepot: use fixed-sized slots for stack records")Signed-off-by Waiman Long <[email protected]>--- lib/stackdepot.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-)lib/stackdepot.c +1
-1@@ -36,11 +36,12 @@#include <linux/memblock.h> #include <linux/kasan-enabled.h> -#define DEPOT_POOLS_CAP 8192 +/* KASAN is a big user of stackdepot, double the cap if KASAN is enabled */
lib/stackdepot.c line 47
Quoted 10 lines
lib/stackdepot.c +3
-2+#define DEPOT_POOLS_CAP (8192 * (IS_ENABLED(CONFIG_KASAN) ? 2 : 1)) +/* The pool_index is offset by 1 so the first record does not have a 0 handle. */ #define DEPOT_MAX_POOLS \ - (((1LL << (DEPOT_POOL_INDEX_BITS)) - 1 < DEPOT_POOLS_CAP) ? \ - (1LL << (DEPOT_POOL_INDEX_BITS)) - 1 : DEPOT_POOLS_CAP) + MIN((1LL << (DEPOT_POOL_INDEX_BITS)) - 1, DEPOT_POOLS_CAP)static bool stack_depot_disabled;static bool __stack_depot_early_init_requested__initdata =IS_ENABLED (CONFIG_STACKDEPOT_ALWAYS_INIT );
Sorry, I accidentally send it twice. Please ignore this version.
BTW, the code change is the same as v1, but the description is updated to reflect my investigation as why we need to double DEPOT_POOLS_CAP for KASAN.
Cheers,
Longman
Quoted 75 lines · Waiman Long
As said in commit
02754e0a484a("lib/stackdepot.c: bump stackdepot capacity from 16MB to 128MB"), KASAN uses stackdepot to memorize stacks for all kmalloc/kfree calls. So stackdepot capacity was increased 8 times to accommodate KASAN usage even thought it was claimed 4X should be enough at that time.With commit
fc60e0caa94d("lib/stackdepot: use fixed-sized slots for stack records"), all stackdepot records uses a fixed size withCONFIG_STACKDEPOT_MAX_FRAMES(default=64) entries. This is merged to support evictable KASAN stack records. Commit31639fd6cebd("stackdepot:
use variable size records for non-evictable entries") re-enabled the use of variable size records for non-KASAN use cases, but KASAN (generic mode) still uses the large fixed size stack records.With the default
CONFIG_STACKDEPOT_MAX_FRAMESof 64, KASAN use of stackdepot space had been more than double than before. Assuming an average stack frame size of 16, a KASAN stack record is almost 4X the size of a non-KASAN one.When a wide variety of workloads are run on a debug kernel with KASAN enabled, the following warning may sometimes be printed.
[ 6818.650674] Stack depot reached limit capacity [ 6818.650730] WARNING: CPU: 1 PID: 272741 at lib/stackdepot.c:252 depot_alloc_stack+0x39e/0x3d0 : [ 6818.650907] Call Trace: [ 6818.650909] [<00047dd453d84b92>] depot_alloc_stack+0x3a2/0x3d0 [ 6818.650916] [<00047dd453d85254>] stack_depot_save_flags+0x4f4/0x5c0 [ 6818.650920] [<00047dd4535872c6>] kasan_save_stack+0x56/0x70 [ 6818.650924] [<00047dd453587328>] kasan_save_track+0x28/0x40 [ 6818.650927] [<00047dd45358a27a>] kasan_save_free_info+0x4a/0x70 [ 6818.650930] [<00047dd45358766a>] __kasan_slab_free+0x12a/0x1d0 [ 6818.650933] [<00047dd45350deb4>] kmem_cache_free+0x1b4/0x580 [ 6818.650938] [<00047dd452c520da>] __put_task_struct+0x24a/0x320 [ 6818.650945] [<00047dd452c6aee4>] delayed_put_task_struct+0x294/0x350 [ 6818.650949] [<00047dd452e9066a>] rcu_do_batch+0x6ea/0x2090 [ 6818.650953] [<00047dd452ea60f4>] rcu_core+0x474/0xa90 [ 6818.650956] [<00047dd452c780c0>] handle_softirqs+0x3c0/0xf90 [ 6818.650960] [<00047dd452c76fbe>] __irq_exit_rcu+0x35e/0x460 [ 6818.650963] [<00047dd452c79992>] irq_exit_rcu+0x22/0xb0 [ 6818.650966] [<00047dd454bd8128>] do_ext_irq+0xd8/0x120 [ 6818.650972] [<00047dd454c0ddd0>] ext_int_handler+0xb8/0xe8 [ 6818.650979] [<00047dd453589cf6>] kasan_check_range+0x236/0x2f0 [ 6818.650982] [<00047dd453378cf0>] filemap_get_pages+0x190/0xaa0 [ 6818.650986] [<00047dd453379940>] filemap_read+0x340/0xa70 [ 6818.650989] [<00047dd3d325d226>] xfs_file_buffered_read+0x2c6/0x400 [xfs] [ 6818.651431] [<00047dd3d325dfe2>] xfs_file_read_iter+0x2c2/0x550 [xfs] [ 6818.651663] [<00047dd45364710c>] vfs_read+0x64c/0x8c0 [ 6818.651669] [<00047dd453648ed8>] ksys_read+0x118/0x200 [ 6818.651672] [<00047dd452b6cf5a>] do_syscall+0x27a/0x380 [ 6818.651676] [<00047dd454bd7e74>] __do_syscall+0xf4/0x1a0 [ 6818.651680] [<00047dd454c0db58>] system_call+0x70/0x98With all the recent changes in stackdepot to support new KASAN features, it is obvious that the current
DEPOT_POOLS_CAPof 8192 may not be enough when KASAN is enabled. Fix this stackdepot capability issue by doublingDEPOT_POOLS_CAPif KASAN is enabled. With 4k pages, the maximum stackdepot capacity is doubled to 256 MB with KASAN enabled.Also use the
MIN()macro for definingDEPOT_MAX_POOLSto clarify the intention.Fixes fc60e0caa94d ("lib/stackdepot: use fixed-sized slots for stack records")Signed-off-by Waiman Long <[email protected]>--- lib/stackdepot.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-)lib/stackdepot.c +0
-0@@ -36,11 +36,12 @@#include <linux/memblock.h> #include <linux/kasan-enabled.h>
lib/stackdepot.c line 44
Quoted 10 lines
lib/stackdepot.c +4
-3-#define DEPOT_POOLS_CAP 8192 +/* KASAN is a big user of stackdepot, double the cap if KASAN is enabled */ +#define DEPOT_POOLS_CAP (8192 * (IS_ENABLED(CONFIG_KASAN) ? 2 : 1)) +/* The pool_index is offset by 1 so the first record does not have a 0 handle. */ #define DEPOT_MAX_POOLS \ - (((1LL << (DEPOT_POOL_INDEX_BITS)) - 1 < DEPOT_POOLS_CAP) ? \ - (1LL << (DEPOT_POOL_INDEX_BITS)) - 1 : DEPOT_POOLS_CAP) + MIN((1LL << (DEPOT_POOL_INDEX_BITS)) - 1, DEPOT_POOLS_CAP)
This breaks the build for the mm-unstable branch in Andrew's mm tree which this patch has been taken to.
This is because we don't appear to have commit 1a251f52cfdc ("minmax: make generic MIN() and MAX() macros available everywhere").
This commit is in Linus's tree so perhaps the easiest solution is to pull this into the mm-unstable branch?
Thanks!
Quoted 7 lines
lib/stackdepot.c +0
-1static bool stack_depot_disabled;static bool __stack_depot_early_init_requested__initdata =IS_ENABLED (CONFIG_STACKDEPOT_ALWAYS_INIT ); --2.43.5