February 2025 - Linux-stable-mirror

[PATCH AUTOSEL 5.4 1/2] orangefs: fix a oob in orangefs_debug_write

by Sasha Levin

From: Mike Marshall <hubcap(a)omnibond.com> [ Upstream commit f7c848431632598ff9bce57a659db6af60d75b39 ] I got a syzbot report: slab-out-of-bounds Read in orangefs_debug_write... several people suggested fixes, I tested Al Viro's suggestion and made this patch. Signed-off-by: Mike Marshall <hubcap(a)omnibond.com> Reported-by: syzbot+fc519d7875f2d9186c1f(a)syzkaller.appspotmail.com Signed-off-by: Sasha Levin <sashal(a)kernel.org> --- fs/orangefs/orangefs-debugfs.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/fs/orangefs/orangefs-debugfs.c b/fs/orangefs/orangefs-debugfs.c index 1b508f5433846..fa41db0884880 100644 --- a/fs/orangefs/orangefs-debugfs.c +++ b/fs/orangefs/orangefs-debugfs.c @@ -393,9 +393,9 @@ static ssize_t orangefs_debug_write(struct file *file, * Thwart users who try to jamb a ridiculous number * of bytes into the debug file... */ - if (count > ORANGEFS_MAX_DEBUG_STRING_LEN + 1) { + if (count > ORANGEFS_MAX_DEBUG_STRING_LEN) { silly = count; - count = ORANGEFS_MAX_DEBUG_STRING_LEN + 1; + count = ORANGEFS_MAX_DEBUG_STRING_LEN; } buf = kzalloc(ORANGEFS_MAX_DEBUG_STRING_LEN, GFP_KERNEL); -- 2.39.5

4 months, 3 weeks

1
1
0 0

[PATCH AUTOSEL 5.10 1/2] orangefs: fix a oob in orangefs_debug_write

by Sasha Levin

From: Mike Marshall <hubcap(a)omnibond.com> [ Upstream commit f7c848431632598ff9bce57a659db6af60d75b39 ] I got a syzbot report: slab-out-of-bounds Read in orangefs_debug_write... several people suggested fixes, I tested Al Viro's suggestion and made this patch. Signed-off-by: Mike Marshall <hubcap(a)omnibond.com> Reported-by: syzbot+fc519d7875f2d9186c1f(a)syzkaller.appspotmail.com Signed-off-by: Sasha Levin <sashal(a)kernel.org> --- fs/orangefs/orangefs-debugfs.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/fs/orangefs/orangefs-debugfs.c b/fs/orangefs/orangefs-debugfs.c index 1b508f5433846..fa41db0884880 100644 --- a/fs/orangefs/orangefs-debugfs.c +++ b/fs/orangefs/orangefs-debugfs.c @@ -393,9 +393,9 @@ static ssize_t orangefs_debug_write(struct file *file, * Thwart users who try to jamb a ridiculous number * of bytes into the debug file... */ - if (count > ORANGEFS_MAX_DEBUG_STRING_LEN + 1) { + if (count > ORANGEFS_MAX_DEBUG_STRING_LEN) { silly = count; - count = ORANGEFS_MAX_DEBUG_STRING_LEN + 1; + count = ORANGEFS_MAX_DEBUG_STRING_LEN; } buf = kzalloc(ORANGEFS_MAX_DEBUG_STRING_LEN, GFP_KERNEL); -- 2.39.5

4 months, 3 weeks

1
1
0 0

[PATCH AUTOSEL 5.15 1/3] x86/mm/tlb: Only trim the mm_cpumask once a second

by Sasha Levin

From: Rik van Riel <riel(a)fb.com> [ Upstream commit 6db2526c1d694c91c6e05e2f186c085e9460f202 ] Setting and clearing CPU bits in the mm_cpumask is only ever done by the CPU itself, from the context switch code or the TLB flush code. Synchronization is handled by switch_mm_irqs_off() blocking interrupts. Sending TLB flush IPIs to CPUs that are in the mm_cpumask, but no longer running the program causes a regression in the will-it-scale tlbflush2 test. This test is contrived, but a large regression here might cause a small regression in some real world workload. Instead of always sending IPIs to CPUs that are in the mm_cpumask, but no longer running the program, send these IPIs only once a second. The rest of the time we can skip over CPUs where the loaded_mm is different from the target mm. Reported-by: kernel test roboto <oliver.sang(a)intel.com> Signed-off-by: Rik van Riel <riel(a)surriel.com> Signed-off-by: Ingo Molnar <mingo(a)kernel.org> Cc: Dave Hansen <dave.hansen(a)linux.intel.com> Cc: Andy Lutomirski <luto(a)kernel.org> Cc: Mathieu Desnoyers <mathieu.desnoyers(a)efficios.com> Cc: Peter Zijlstra <peterz(a)infradead.org> Cc: Linus Torvalds <torvalds(a)linux-foundation.org> Link: https://lore.kernel.org/r/20241204210316.612ee573@fangorn Closes: https://lore.kernel.org/oe-lkp/202411282207.6bd28eae-lkp@intel.com/ Signed-off-by: Sasha Levin <sashal(a)kernel.org> --- arch/x86/include/asm/mmu.h | 2 ++ arch/x86/include/asm/mmu_context.h | 1 + arch/x86/include/asm/tlbflush.h | 1 + arch/x86/mm/tlb.c | 35 +++++++++++++++++++++++++++--- 4 files changed, 36 insertions(+), 3 deletions(-) diff --git a/arch/x86/include/asm/mmu.h b/arch/x86/include/asm/mmu.h index 5d7494631ea95..c07c018a1c139 100644 --- a/arch/x86/include/asm/mmu.h +++ b/arch/x86/include/asm/mmu.h @@ -33,6 +33,8 @@ typedef struct { */ atomic64_t tlb_gen; + unsigned long next_trim_cpumask; + #ifdef CONFIG_MODIFY_LDT_SYSCALL struct rw_semaphore ldt_usr_sem; struct ldt_struct *ldt; diff --git a/arch/x86/include/asm/mmu_context.h b/arch/x86/include/asm/mmu_context.h index 27516046117a3..1d11aeb597542 100644 --- a/arch/x86/include/asm/mmu_context.h +++ b/arch/x86/include/asm/mmu_context.h @@ -106,6 +106,7 @@ static inline int init_new_context(struct task_struct *tsk, mm->context.ctx_id = atomic64_inc_return(&last_mm_ctx_id); atomic64_set(&mm->context.tlb_gen, 0); + mm->context.next_trim_cpumask = jiffies + HZ; #ifdef CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS if (cpu_feature_enabled(X86_FEATURE_OSPKE)) { diff --git a/arch/x86/include/asm/tlbflush.h b/arch/x86/include/asm/tlbflush.h index b587a9ee9cb25..22b93e35fa886 100644 --- a/arch/x86/include/asm/tlbflush.h +++ b/arch/x86/include/asm/tlbflush.h @@ -207,6 +207,7 @@ struct flush_tlb_info { unsigned int initiating_cpu; u8 stride_shift; u8 freed_tables; + u8 trim_cpumask; }; void flush_tlb_local(void); diff --git a/arch/x86/mm/tlb.c b/arch/x86/mm/tlb.c index 511172d70825c..19d083ad2de79 100644 --- a/arch/x86/mm/tlb.c +++ b/arch/x86/mm/tlb.c @@ -854,9 +854,36 @@ static void flush_tlb_func(void *info) nr_invalidate); } -static bool tlb_is_not_lazy(int cpu, void *data) +static bool should_flush_tlb(int cpu, void *data) { - return !per_cpu(cpu_tlbstate_shared.is_lazy, cpu); + struct flush_tlb_info *info = data; + + /* Lazy TLB will get flushed at the next context switch. */ + if (per_cpu(cpu_tlbstate_shared.is_lazy, cpu)) + return false; + + /* No mm means kernel memory flush. */ + if (!info->mm) + return true; + + /* The target mm is loaded, and the CPU is not lazy. */ + if (per_cpu(cpu_tlbstate.loaded_mm, cpu) == info->mm) + return true; + + /* In cpumask, but not the loaded mm? Periodically remove by flushing. */ + if (info->trim_cpumask) + return true; + + return false; +} + +static bool should_trim_cpumask(struct mm_struct *mm) +{ + if (time_after(jiffies, READ_ONCE(mm->context.next_trim_cpumask))) { + WRITE_ONCE(mm->context.next_trim_cpumask, jiffies + HZ); + return true; + } + return false; } DEFINE_PER_CPU_SHARED_ALIGNED(struct tlb_state_shared, cpu_tlbstate_shared); @@ -890,7 +917,7 @@ STATIC_NOPV void native_flush_tlb_multi(const struct cpumask *cpumask, if (info->freed_tables) on_each_cpu_mask(cpumask, flush_tlb_func, (void *)info, true); else - on_each_cpu_cond_mask(tlb_is_not_lazy, flush_tlb_func, + on_each_cpu_cond_mask(should_flush_tlb, flush_tlb_func, (void *)info, 1, cpumask); } @@ -941,6 +968,7 @@ static struct flush_tlb_info *get_flush_tlb_info(struct mm_struct *mm, info->freed_tables = freed_tables; info->new_tlb_gen = new_tlb_gen; info->initiating_cpu = smp_processor_id(); + info->trim_cpumask = 0; return info; } @@ -983,6 +1011,7 @@ void flush_tlb_mm_range(struct mm_struct *mm, unsigned long start, * flush_tlb_func_local() directly in this case. */ if (cpumask_any_but(mm_cpumask(mm), cpu) < nr_cpu_ids) { + info->trim_cpumask = should_trim_cpumask(mm); flush_tlb_multi(mm_cpumask(mm), info); } else if (mm == this_cpu_read(cpu_tlbstate.loaded_mm)) { lockdep_assert_irqs_enabled(); -- 2.39.5

4 months, 3 weeks

1
2
0 0

[PATCH AUTOSEL 6.1 1/3] x86/mm/tlb: Only trim the mm_cpumask once a second

by Sasha Levin

From: Rik van Riel <riel(a)fb.com> [ Upstream commit 6db2526c1d694c91c6e05e2f186c085e9460f202 ] Setting and clearing CPU bits in the mm_cpumask is only ever done by the CPU itself, from the context switch code or the TLB flush code. Synchronization is handled by switch_mm_irqs_off() blocking interrupts. Sending TLB flush IPIs to CPUs that are in the mm_cpumask, but no longer running the program causes a regression in the will-it-scale tlbflush2 test. This test is contrived, but a large regression here might cause a small regression in some real world workload. Instead of always sending IPIs to CPUs that are in the mm_cpumask, but no longer running the program, send these IPIs only once a second. The rest of the time we can skip over CPUs where the loaded_mm is different from the target mm. Reported-by: kernel test roboto <oliver.sang(a)intel.com> Signed-off-by: Rik van Riel <riel(a)surriel.com> Signed-off-by: Ingo Molnar <mingo(a)kernel.org> Cc: Dave Hansen <dave.hansen(a)linux.intel.com> Cc: Andy Lutomirski <luto(a)kernel.org> Cc: Mathieu Desnoyers <mathieu.desnoyers(a)efficios.com> Cc: Peter Zijlstra <peterz(a)infradead.org> Cc: Linus Torvalds <torvalds(a)linux-foundation.org> Link: https://lore.kernel.org/r/20241204210316.612ee573@fangorn Closes: https://lore.kernel.org/oe-lkp/202411282207.6bd28eae-lkp@intel.com/ Signed-off-by: Sasha Levin <sashal(a)kernel.org> --- arch/x86/include/asm/mmu.h | 2 ++ arch/x86/include/asm/mmu_context.h | 1 + arch/x86/include/asm/tlbflush.h | 1 + arch/x86/mm/tlb.c | 35 +++++++++++++++++++++++++++--- 4 files changed, 36 insertions(+), 3 deletions(-) diff --git a/arch/x86/include/asm/mmu.h b/arch/x86/include/asm/mmu.h index 5d7494631ea95..c07c018a1c139 100644 --- a/arch/x86/include/asm/mmu.h +++ b/arch/x86/include/asm/mmu.h @@ -33,6 +33,8 @@ typedef struct { */ atomic64_t tlb_gen; + unsigned long next_trim_cpumask; + #ifdef CONFIG_MODIFY_LDT_SYSCALL struct rw_semaphore ldt_usr_sem; struct ldt_struct *ldt; diff --git a/arch/x86/include/asm/mmu_context.h b/arch/x86/include/asm/mmu_context.h index b8d40ddeab00f..6f5c7584fe1e3 100644 --- a/arch/x86/include/asm/mmu_context.h +++ b/arch/x86/include/asm/mmu_context.h @@ -106,6 +106,7 @@ static inline int init_new_context(struct task_struct *tsk, mm->context.ctx_id = atomic64_inc_return(&last_mm_ctx_id); atomic64_set(&mm->context.tlb_gen, 0); + mm->context.next_trim_cpumask = jiffies + HZ; #ifdef CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS if (cpu_feature_enabled(X86_FEATURE_OSPKE)) { diff --git a/arch/x86/include/asm/tlbflush.h b/arch/x86/include/asm/tlbflush.h index cda3118f3b27d..d1eb6bbfd39e3 100644 --- a/arch/x86/include/asm/tlbflush.h +++ b/arch/x86/include/asm/tlbflush.h @@ -208,6 +208,7 @@ struct flush_tlb_info { unsigned int initiating_cpu; u8 stride_shift; u8 freed_tables; + u8 trim_cpumask; }; void flush_tlb_local(void); diff --git a/arch/x86/mm/tlb.c b/arch/x86/mm/tlb.c index c1e31e9a85d76..b07e2167fcebf 100644 --- a/arch/x86/mm/tlb.c +++ b/arch/x86/mm/tlb.c @@ -878,9 +878,36 @@ static void flush_tlb_func(void *info) nr_invalidate); } -static bool tlb_is_not_lazy(int cpu, void *data) +static bool should_flush_tlb(int cpu, void *data) { - return !per_cpu(cpu_tlbstate_shared.is_lazy, cpu); + struct flush_tlb_info *info = data; + + /* Lazy TLB will get flushed at the next context switch. */ + if (per_cpu(cpu_tlbstate_shared.is_lazy, cpu)) + return false; + + /* No mm means kernel memory flush. */ + if (!info->mm) + return true; + + /* The target mm is loaded, and the CPU is not lazy. */ + if (per_cpu(cpu_tlbstate.loaded_mm, cpu) == info->mm) + return true; + + /* In cpumask, but not the loaded mm? Periodically remove by flushing. */ + if (info->trim_cpumask) + return true; + + return false; +} + +static bool should_trim_cpumask(struct mm_struct *mm) +{ + if (time_after(jiffies, READ_ONCE(mm->context.next_trim_cpumask))) { + WRITE_ONCE(mm->context.next_trim_cpumask, jiffies + HZ); + return true; + } + return false; } DEFINE_PER_CPU_SHARED_ALIGNED(struct tlb_state_shared, cpu_tlbstate_shared); @@ -914,7 +941,7 @@ STATIC_NOPV void native_flush_tlb_multi(const struct cpumask *cpumask, if (info->freed_tables) on_each_cpu_mask(cpumask, flush_tlb_func, (void *)info, true); else - on_each_cpu_cond_mask(tlb_is_not_lazy, flush_tlb_func, + on_each_cpu_cond_mask(should_flush_tlb, flush_tlb_func, (void *)info, 1, cpumask); } @@ -965,6 +992,7 @@ static struct flush_tlb_info *get_flush_tlb_info(struct mm_struct *mm, info->freed_tables = freed_tables; info->new_tlb_gen = new_tlb_gen; info->initiating_cpu = smp_processor_id(); + info->trim_cpumask = 0; return info; } @@ -1007,6 +1035,7 @@ void flush_tlb_mm_range(struct mm_struct *mm, unsigned long start, * flush_tlb_func_local() directly in this case. */ if (cpumask_any_but(mm_cpumask(mm), cpu) < nr_cpu_ids) { + info->trim_cpumask = should_trim_cpumask(mm); flush_tlb_multi(mm_cpumask(mm), info); } else if (mm == this_cpu_read(cpu_tlbstate.loaded_mm)) { lockdep_assert_irqs_enabled(); -- 2.39.5

4 months, 3 weeks

1
2
0 0

[PATCH AUTOSEL 6.6 1/3] x86/mm/tlb: Only trim the mm_cpumask once a second

by Sasha Levin

From: Rik van Riel <riel(a)fb.com> [ Upstream commit 6db2526c1d694c91c6e05e2f186c085e9460f202 ] Setting and clearing CPU bits in the mm_cpumask is only ever done by the CPU itself, from the context switch code or the TLB flush code. Synchronization is handled by switch_mm_irqs_off() blocking interrupts. Sending TLB flush IPIs to CPUs that are in the mm_cpumask, but no longer running the program causes a regression in the will-it-scale tlbflush2 test. This test is contrived, but a large regression here might cause a small regression in some real world workload. Instead of always sending IPIs to CPUs that are in the mm_cpumask, but no longer running the program, send these IPIs only once a second. The rest of the time we can skip over CPUs where the loaded_mm is different from the target mm. Reported-by: kernel test roboto <oliver.sang(a)intel.com> Signed-off-by: Rik van Riel <riel(a)surriel.com> Signed-off-by: Ingo Molnar <mingo(a)kernel.org> Cc: Dave Hansen <dave.hansen(a)linux.intel.com> Cc: Andy Lutomirski <luto(a)kernel.org> Cc: Mathieu Desnoyers <mathieu.desnoyers(a)efficios.com> Cc: Peter Zijlstra <peterz(a)infradead.org> Cc: Linus Torvalds <torvalds(a)linux-foundation.org> Link: https://lore.kernel.org/r/20241204210316.612ee573@fangorn Closes: https://lore.kernel.org/oe-lkp/202411282207.6bd28eae-lkp@intel.com/ Signed-off-by: Sasha Levin <sashal(a)kernel.org> --- arch/x86/include/asm/mmu.h | 2 ++ arch/x86/include/asm/mmu_context.h | 1 + arch/x86/include/asm/tlbflush.h | 1 + arch/x86/mm/tlb.c | 35 +++++++++++++++++++++++++++--- 4 files changed, 36 insertions(+), 3 deletions(-) diff --git a/arch/x86/include/asm/mmu.h b/arch/x86/include/asm/mmu.h index 0da5c227f490c..53763cf192777 100644 --- a/arch/x86/include/asm/mmu.h +++ b/arch/x86/include/asm/mmu.h @@ -37,6 +37,8 @@ typedef struct { */ atomic64_t tlb_gen; + unsigned long next_trim_cpumask; + #ifdef CONFIG_MODIFY_LDT_SYSCALL struct rw_semaphore ldt_usr_sem; struct ldt_struct *ldt; diff --git a/arch/x86/include/asm/mmu_context.h b/arch/x86/include/asm/mmu_context.h index 8dac45a2c7fcf..f5afd956d5e50 100644 --- a/arch/x86/include/asm/mmu_context.h +++ b/arch/x86/include/asm/mmu_context.h @@ -145,6 +145,7 @@ static inline int init_new_context(struct task_struct *tsk, mm->context.ctx_id = atomic64_inc_return(&last_mm_ctx_id); atomic64_set(&mm->context.tlb_gen, 0); + mm->context.next_trim_cpumask = jiffies + HZ; #ifdef CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS if (cpu_feature_enabled(X86_FEATURE_OSPKE)) { diff --git a/arch/x86/include/asm/tlbflush.h b/arch/x86/include/asm/tlbflush.h index 25726893c6f4d..5d61adc6e892e 100644 --- a/arch/x86/include/asm/tlbflush.h +++ b/arch/x86/include/asm/tlbflush.h @@ -222,6 +222,7 @@ struct flush_tlb_info { unsigned int initiating_cpu; u8 stride_shift; u8 freed_tables; + u8 trim_cpumask; }; void flush_tlb_local(void); diff --git a/arch/x86/mm/tlb.c b/arch/x86/mm/tlb.c index 64f594826a282..df1794a5e38a5 100644 --- a/arch/x86/mm/tlb.c +++ b/arch/x86/mm/tlb.c @@ -898,9 +898,36 @@ static void flush_tlb_func(void *info) nr_invalidate); } -static bool tlb_is_not_lazy(int cpu, void *data) +static bool should_flush_tlb(int cpu, void *data) { - return !per_cpu(cpu_tlbstate_shared.is_lazy, cpu); + struct flush_tlb_info *info = data; + + /* Lazy TLB will get flushed at the next context switch. */ + if (per_cpu(cpu_tlbstate_shared.is_lazy, cpu)) + return false; + + /* No mm means kernel memory flush. */ + if (!info->mm) + return true; + + /* The target mm is loaded, and the CPU is not lazy. */ + if (per_cpu(cpu_tlbstate.loaded_mm, cpu) == info->mm) + return true; + + /* In cpumask, but not the loaded mm? Periodically remove by flushing. */ + if (info->trim_cpumask) + return true; + + return false; +} + +static bool should_trim_cpumask(struct mm_struct *mm) +{ + if (time_after(jiffies, READ_ONCE(mm->context.next_trim_cpumask))) { + WRITE_ONCE(mm->context.next_trim_cpumask, jiffies + HZ); + return true; + } + return false; } DEFINE_PER_CPU_SHARED_ALIGNED(struct tlb_state_shared, cpu_tlbstate_shared); @@ -934,7 +961,7 @@ STATIC_NOPV void native_flush_tlb_multi(const struct cpumask *cpumask, if (info->freed_tables) on_each_cpu_mask(cpumask, flush_tlb_func, (void *)info, true); else - on_each_cpu_cond_mask(tlb_is_not_lazy, flush_tlb_func, + on_each_cpu_cond_mask(should_flush_tlb, flush_tlb_func, (void *)info, 1, cpumask); } @@ -985,6 +1012,7 @@ static struct flush_tlb_info *get_flush_tlb_info(struct mm_struct *mm, info->freed_tables = freed_tables; info->new_tlb_gen = new_tlb_gen; info->initiating_cpu = smp_processor_id(); + info->trim_cpumask = 0; return info; } @@ -1027,6 +1055,7 @@ void flush_tlb_mm_range(struct mm_struct *mm, unsigned long start, * flush_tlb_func_local() directly in this case. */ if (cpumask_any_but(mm_cpumask(mm), cpu) < nr_cpu_ids) { + info->trim_cpumask = should_trim_cpumask(mm); flush_tlb_multi(mm_cpumask(mm), info); } else if (mm == this_cpu_read(cpu_tlbstate.loaded_mm)) { lockdep_assert_irqs_enabled(); -- 2.39.5

4 months, 3 weeks

1
2
0 0

[PATCH AUTOSEL 6.12 1/5] x86/mm/tlb: Only trim the mm_cpumask once a second

by Sasha Levin

From: Rik van Riel <riel(a)fb.com> [ Upstream commit 6db2526c1d694c91c6e05e2f186c085e9460f202 ] Setting and clearing CPU bits in the mm_cpumask is only ever done by the CPU itself, from the context switch code or the TLB flush code. Synchronization is handled by switch_mm_irqs_off() blocking interrupts. Sending TLB flush IPIs to CPUs that are in the mm_cpumask, but no longer running the program causes a regression in the will-it-scale tlbflush2 test. This test is contrived, but a large regression here might cause a small regression in some real world workload. Instead of always sending IPIs to CPUs that are in the mm_cpumask, but no longer running the program, send these IPIs only once a second. The rest of the time we can skip over CPUs where the loaded_mm is different from the target mm. Reported-by: kernel test roboto <oliver.sang(a)intel.com> Signed-off-by: Rik van Riel <riel(a)surriel.com> Signed-off-by: Ingo Molnar <mingo(a)kernel.org> Cc: Dave Hansen <dave.hansen(a)linux.intel.com> Cc: Andy Lutomirski <luto(a)kernel.org> Cc: Mathieu Desnoyers <mathieu.desnoyers(a)efficios.com> Cc: Peter Zijlstra <peterz(a)infradead.org> Cc: Linus Torvalds <torvalds(a)linux-foundation.org> Link: https://lore.kernel.org/r/20241204210316.612ee573@fangorn Closes: https://lore.kernel.org/oe-lkp/202411282207.6bd28eae-lkp@intel.com/ Signed-off-by: Sasha Levin <sashal(a)kernel.org> --- arch/x86/include/asm/mmu.h | 2 ++ arch/x86/include/asm/mmu_context.h | 1 + arch/x86/include/asm/tlbflush.h | 1 + arch/x86/mm/tlb.c | 35 +++++++++++++++++++++++++++--- 4 files changed, 36 insertions(+), 3 deletions(-) diff --git a/arch/x86/include/asm/mmu.h b/arch/x86/include/asm/mmu.h index ce4677b8b7356..3b496cdcb74b3 100644 --- a/arch/x86/include/asm/mmu.h +++ b/arch/x86/include/asm/mmu.h @@ -37,6 +37,8 @@ typedef struct { */ atomic64_t tlb_gen; + unsigned long next_trim_cpumask; + #ifdef CONFIG_MODIFY_LDT_SYSCALL struct rw_semaphore ldt_usr_sem; struct ldt_struct *ldt; diff --git a/arch/x86/include/asm/mmu_context.h b/arch/x86/include/asm/mmu_context.h index 2886cb668d7fa..795fdd53bd0a6 100644 --- a/arch/x86/include/asm/mmu_context.h +++ b/arch/x86/include/asm/mmu_context.h @@ -151,6 +151,7 @@ static inline int init_new_context(struct task_struct *tsk, mm->context.ctx_id = atomic64_inc_return(&last_mm_ctx_id); atomic64_set(&mm->context.tlb_gen, 0); + mm->context.next_trim_cpumask = jiffies + HZ; #ifdef CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS if (cpu_feature_enabled(X86_FEATURE_OSPKE)) { diff --git a/arch/x86/include/asm/tlbflush.h b/arch/x86/include/asm/tlbflush.h index 69e79fff41b80..02fc2aa06e9e0 100644 --- a/arch/x86/include/asm/tlbflush.h +++ b/arch/x86/include/asm/tlbflush.h @@ -222,6 +222,7 @@ struct flush_tlb_info { unsigned int initiating_cpu; u8 stride_shift; u8 freed_tables; + u8 trim_cpumask; }; void flush_tlb_local(void); diff --git a/arch/x86/mm/tlb.c b/arch/x86/mm/tlb.c index b0678d59ebdb4..00ffa74d0dd0b 100644 --- a/arch/x86/mm/tlb.c +++ b/arch/x86/mm/tlb.c @@ -893,9 +893,36 @@ static void flush_tlb_func(void *info) nr_invalidate); } -static bool tlb_is_not_lazy(int cpu, void *data) +static bool should_flush_tlb(int cpu, void *data) { - return !per_cpu(cpu_tlbstate_shared.is_lazy, cpu); + struct flush_tlb_info *info = data; + + /* Lazy TLB will get flushed at the next context switch. */ + if (per_cpu(cpu_tlbstate_shared.is_lazy, cpu)) + return false; + + /* No mm means kernel memory flush. */ + if (!info->mm) + return true; + + /* The target mm is loaded, and the CPU is not lazy. */ + if (per_cpu(cpu_tlbstate.loaded_mm, cpu) == info->mm) + return true; + + /* In cpumask, but not the loaded mm? Periodically remove by flushing. */ + if (info->trim_cpumask) + return true; + + return false; +} + +static bool should_trim_cpumask(struct mm_struct *mm) +{ + if (time_after(jiffies, READ_ONCE(mm->context.next_trim_cpumask))) { + WRITE_ONCE(mm->context.next_trim_cpumask, jiffies + HZ); + return true; + } + return false; } DEFINE_PER_CPU_SHARED_ALIGNED(struct tlb_state_shared, cpu_tlbstate_shared); @@ -929,7 +956,7 @@ STATIC_NOPV void native_flush_tlb_multi(const struct cpumask *cpumask, if (info->freed_tables) on_each_cpu_mask(cpumask, flush_tlb_func, (void *)info, true); else - on_each_cpu_cond_mask(tlb_is_not_lazy, flush_tlb_func, + on_each_cpu_cond_mask(should_flush_tlb, flush_tlb_func, (void *)info, 1, cpumask); } @@ -980,6 +1007,7 @@ static struct flush_tlb_info *get_flush_tlb_info(struct mm_struct *mm, info->freed_tables = freed_tables; info->new_tlb_gen = new_tlb_gen; info->initiating_cpu = smp_processor_id(); + info->trim_cpumask = 0; return info; } @@ -1022,6 +1050,7 @@ void flush_tlb_mm_range(struct mm_struct *mm, unsigned long start, * flush_tlb_func_local() directly in this case. */ if (cpumask_any_but(mm_cpumask(mm), cpu) < nr_cpu_ids) { + info->trim_cpumask = should_trim_cpumask(mm); flush_tlb_multi(mm_cpumask(mm), info); } else if (mm == this_cpu_read(cpu_tlbstate.loaded_mm)) { lockdep_assert_irqs_enabled(); -- 2.39.5

4 months, 3 weeks

1
4
0 0

[PATCH AUTOSEL 6.13 1/6] x86/mm/tlb: Only trim the mm_cpumask once a second

by Sasha Levin

From: Rik van Riel <riel(a)fb.com> [ Upstream commit 6db2526c1d694c91c6e05e2f186c085e9460f202 ] Setting and clearing CPU bits in the mm_cpumask is only ever done by the CPU itself, from the context switch code or the TLB flush code. Synchronization is handled by switch_mm_irqs_off() blocking interrupts. Sending TLB flush IPIs to CPUs that are in the mm_cpumask, but no longer running the program causes a regression in the will-it-scale tlbflush2 test. This test is contrived, but a large regression here might cause a small regression in some real world workload. Instead of always sending IPIs to CPUs that are in the mm_cpumask, but no longer running the program, send these IPIs only once a second. The rest of the time we can skip over CPUs where the loaded_mm is different from the target mm. Reported-by: kernel test roboto <oliver.sang(a)intel.com> Signed-off-by: Rik van Riel <riel(a)surriel.com> Signed-off-by: Ingo Molnar <mingo(a)kernel.org> Cc: Dave Hansen <dave.hansen(a)linux.intel.com> Cc: Andy Lutomirski <luto(a)kernel.org> Cc: Mathieu Desnoyers <mathieu.desnoyers(a)efficios.com> Cc: Peter Zijlstra <peterz(a)infradead.org> Cc: Linus Torvalds <torvalds(a)linux-foundation.org> Link: https://lore.kernel.org/r/20241204210316.612ee573@fangorn Closes: https://lore.kernel.org/oe-lkp/202411282207.6bd28eae-lkp@intel.com/ Signed-off-by: Sasha Levin <sashal(a)kernel.org> --- arch/x86/include/asm/mmu.h | 2 ++ arch/x86/include/asm/mmu_context.h | 1 + arch/x86/include/asm/tlbflush.h | 1 + arch/x86/mm/tlb.c | 35 +++++++++++++++++++++++++++--- 4 files changed, 36 insertions(+), 3 deletions(-) diff --git a/arch/x86/include/asm/mmu.h b/arch/x86/include/asm/mmu.h index ce4677b8b7356..3b496cdcb74b3 100644 --- a/arch/x86/include/asm/mmu.h +++ b/arch/x86/include/asm/mmu.h @@ -37,6 +37,8 @@ typedef struct { */ atomic64_t tlb_gen; + unsigned long next_trim_cpumask; + #ifdef CONFIG_MODIFY_LDT_SYSCALL struct rw_semaphore ldt_usr_sem; struct ldt_struct *ldt; diff --git a/arch/x86/include/asm/mmu_context.h b/arch/x86/include/asm/mmu_context.h index 2886cb668d7fa..795fdd53bd0a6 100644 --- a/arch/x86/include/asm/mmu_context.h +++ b/arch/x86/include/asm/mmu_context.h @@ -151,6 +151,7 @@ static inline int init_new_context(struct task_struct *tsk, mm->context.ctx_id = atomic64_inc_return(&last_mm_ctx_id); atomic64_set(&mm->context.tlb_gen, 0); + mm->context.next_trim_cpumask = jiffies + HZ; #ifdef CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS if (cpu_feature_enabled(X86_FEATURE_OSPKE)) { diff --git a/arch/x86/include/asm/tlbflush.h b/arch/x86/include/asm/tlbflush.h index 69e79fff41b80..02fc2aa06e9e0 100644 --- a/arch/x86/include/asm/tlbflush.h +++ b/arch/x86/include/asm/tlbflush.h @@ -222,6 +222,7 @@ struct flush_tlb_info { unsigned int initiating_cpu; u8 stride_shift; u8 freed_tables; + u8 trim_cpumask; }; void flush_tlb_local(void); diff --git a/arch/x86/mm/tlb.c b/arch/x86/mm/tlb.c index a2becb85bea79..90a9e47409131 100644 --- a/arch/x86/mm/tlb.c +++ b/arch/x86/mm/tlb.c @@ -893,9 +893,36 @@ static void flush_tlb_func(void *info) nr_invalidate); } -static bool tlb_is_not_lazy(int cpu, void *data) +static bool should_flush_tlb(int cpu, void *data) { - return !per_cpu(cpu_tlbstate_shared.is_lazy, cpu); + struct flush_tlb_info *info = data; + + /* Lazy TLB will get flushed at the next context switch. */ + if (per_cpu(cpu_tlbstate_shared.is_lazy, cpu)) + return false; + + /* No mm means kernel memory flush. */ + if (!info->mm) + return true; + + /* The target mm is loaded, and the CPU is not lazy. */ + if (per_cpu(cpu_tlbstate.loaded_mm, cpu) == info->mm) + return true; + + /* In cpumask, but not the loaded mm? Periodically remove by flushing. */ + if (info->trim_cpumask) + return true; + + return false; +} + +static bool should_trim_cpumask(struct mm_struct *mm) +{ + if (time_after(jiffies, READ_ONCE(mm->context.next_trim_cpumask))) { + WRITE_ONCE(mm->context.next_trim_cpumask, jiffies + HZ); + return true; + } + return false; } DEFINE_PER_CPU_SHARED_ALIGNED(struct tlb_state_shared, cpu_tlbstate_shared); @@ -929,7 +956,7 @@ STATIC_NOPV void native_flush_tlb_multi(const struct cpumask *cpumask, if (info->freed_tables) on_each_cpu_mask(cpumask, flush_tlb_func, (void *)info, true); else - on_each_cpu_cond_mask(tlb_is_not_lazy, flush_tlb_func, + on_each_cpu_cond_mask(should_flush_tlb, flush_tlb_func, (void *)info, 1, cpumask); } @@ -980,6 +1007,7 @@ static struct flush_tlb_info *get_flush_tlb_info(struct mm_struct *mm, info->freed_tables = freed_tables; info->new_tlb_gen = new_tlb_gen; info->initiating_cpu = smp_processor_id(); + info->trim_cpumask = 0; return info; } @@ -1022,6 +1050,7 @@ void flush_tlb_mm_range(struct mm_struct *mm, unsigned long start, * flush_tlb_func_local() directly in this case. */ if (cpumask_any_but(mm_cpumask(mm), cpu) < nr_cpu_ids) { + info->trim_cpumask = should_trim_cpumask(mm); flush_tlb_multi(mm_cpumask(mm), info); } else if (mm == this_cpu_read(cpu_tlbstate.loaded_mm)) { lockdep_assert_irqs_enabled(); -- 2.39.5

4 months, 3 weeks

1
5
0 0

[PATCH AUTOSEL 5.4] Grab mm lock before grabbing pt lock

by Sasha Levin

From: Maksym Planeta <maksym(a)exostellar.io> [ Upstream commit 6d002348789bc16e9203e9818b7a3688787e3b29 ] Function xen_pin_page calls xen_pte_lock, which in turn grab page table lock (ptlock). When locking, xen_pte_lock expect mm->page_table_lock to be held before grabbing ptlock, but this does not happen when pinning is caused by xen_mm_pin_all. This commit addresses lockdep warning below, which shows up when suspending a Xen VM. [ 3680.658422] Freezing user space processes [ 3680.660156] Freezing user space processes completed (elapsed 0.001 seconds) [ 3680.660182] OOM killer disabled. [ 3680.660192] Freezing remaining freezable tasks [ 3680.661485] Freezing remaining freezable tasks completed (elapsed 0.001 seconds) [ 3680.685254] [ 3680.685265] ================================== [ 3680.685269] WARNING: Nested lock was not taken [ 3680.685274] 6.12.0+ #16 Tainted: G W [ 3680.685279] ---------------------------------- [ 3680.685283] migration/0/19 is trying to lock: [ 3680.685288] ffff88800bac33c0 (ptlock_ptr(ptdesc)#2){+.+.}-{3:3}, at: xen_pin_page+0x175/0x1d0 [ 3680.685303] [ 3680.685303] but this task is not holding: [ 3680.685308] init_mm.page_table_lock [ 3680.685311] [ 3680.685311] stack backtrace: [ 3680.685316] CPU: 0 UID: 0 PID: 19 Comm: migration/0 Tainted: G W 6.12.0+ #16 [ 3680.685324] Tainted: [W]=WARN [ 3680.685328] Stopper: multi_cpu_stop+0x0/0x120 <- __stop_cpus.constprop.0+0x8c/0xd0 [ 3680.685339] Call Trace: [ 3680.685344] <TASK> [ 3680.685347] dump_stack_lvl+0x77/0xb0 [ 3680.685356] __lock_acquire+0x917/0x2310 [ 3680.685364] lock_acquire+0xce/0x2c0 [ 3680.685369] ? xen_pin_page+0x175/0x1d0 [ 3680.685373] _raw_spin_lock_nest_lock+0x2f/0x70 [ 3680.685381] ? xen_pin_page+0x175/0x1d0 [ 3680.685386] xen_pin_page+0x175/0x1d0 [ 3680.685390] ? __pfx_xen_pin_page+0x10/0x10 [ 3680.685394] __xen_pgd_walk+0x233/0x2c0 [ 3680.685401] ? stop_one_cpu+0x91/0x100 [ 3680.685405] __xen_pgd_pin+0x5d/0x250 [ 3680.685410] xen_mm_pin_all+0x70/0xa0 [ 3680.685415] xen_pv_pre_suspend+0xf/0x280 [ 3680.685420] xen_suspend+0x57/0x1a0 [ 3680.685428] multi_cpu_stop+0x6b/0x120 [ 3680.685432] ? update_cpumasks_hier+0x7c/0xa60 [ 3680.685439] ? __pfx_multi_cpu_stop+0x10/0x10 [ 3680.685443] cpu_stopper_thread+0x8c/0x140 [ 3680.685448] ? smpboot_thread_fn+0x20/0x1f0 [ 3680.685454] ? __pfx_smpboot_thread_fn+0x10/0x10 [ 3680.685458] smpboot_thread_fn+0xed/0x1f0 [ 3680.685462] kthread+0xde/0x110 [ 3680.685467] ? __pfx_kthread+0x10/0x10 [ 3680.685471] ret_from_fork+0x2f/0x50 [ 3680.685478] ? __pfx_kthread+0x10/0x10 [ 3680.685482] ret_from_fork_asm+0x1a/0x30 [ 3680.685489] </TASK> [ 3680.685491] [ 3680.685491] other info that might help us debug this: [ 3680.685497] 1 lock held by migration/0/19: [ 3680.685500] #0: ffffffff8284df38 (pgd_lock){+.+.}-{3:3}, at: xen_mm_pin_all+0x14/0xa0 [ 3680.685512] [ 3680.685512] stack backtrace: [ 3680.685518] CPU: 0 UID: 0 PID: 19 Comm: migration/0 Tainted: G W 6.12.0+ #16 [ 3680.685528] Tainted: [W]=WARN [ 3680.685531] Stopper: multi_cpu_stop+0x0/0x120 <- __stop_cpus.constprop.0+0x8c/0xd0 [ 3680.685538] Call Trace: [ 3680.685541] <TASK> [ 3680.685544] dump_stack_lvl+0x77/0xb0 [ 3680.685549] __lock_acquire+0x93c/0x2310 [ 3680.685554] lock_acquire+0xce/0x2c0 [ 3680.685558] ? xen_pin_page+0x175/0x1d0 [ 3680.685562] _raw_spin_lock_nest_lock+0x2f/0x70 [ 3680.685568] ? xen_pin_page+0x175/0x1d0 [ 3680.685572] xen_pin_page+0x175/0x1d0 [ 3680.685578] ? __pfx_xen_pin_page+0x10/0x10 [ 3680.685582] __xen_pgd_walk+0x233/0x2c0 [ 3680.685588] ? stop_one_cpu+0x91/0x100 [ 3680.685592] __xen_pgd_pin+0x5d/0x250 [ 3680.685596] xen_mm_pin_all+0x70/0xa0 [ 3680.685600] xen_pv_pre_suspend+0xf/0x280 [ 3680.685607] xen_suspend+0x57/0x1a0 [ 3680.685611] multi_cpu_stop+0x6b/0x120 [ 3680.685615] ? update_cpumasks_hier+0x7c/0xa60 [ 3680.685620] ? __pfx_multi_cpu_stop+0x10/0x10 [ 3680.685625] cpu_stopper_thread+0x8c/0x140 [ 3680.685629] ? smpboot_thread_fn+0x20/0x1f0 [ 3680.685634] ? __pfx_smpboot_thread_fn+0x10/0x10 [ 3680.685638] smpboot_thread_fn+0xed/0x1f0 [ 3680.685642] kthread+0xde/0x110 [ 3680.685645] ? __pfx_kthread+0x10/0x10 [ 3680.685649] ret_from_fork+0x2f/0x50 [ 3680.685654] ? __pfx_kthread+0x10/0x10 [ 3680.685657] ret_from_fork_asm+0x1a/0x30 [ 3680.685662] </TASK> [ 3680.685267] xen:grant_table: Grant tables using version 1 layout [ 3680.685921] OOM killer enabled. [ 3680.685934] Restarting tasks ... done. Signed-off-by: Maksym Planeta <maksym(a)exostellar.io> Reviewed-by: Juergen Gross <jgross(a)suse.com> Message-ID: <20241204103516.3309112-1-maksym(a)exostellar.io> Signed-off-by: Juergen Gross <jgross(a)suse.com> Signed-off-by: Sasha Levin <sashal(a)kernel.org> --- arch/x86/xen/mmu_pv.c | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/arch/x86/xen/mmu_pv.c b/arch/x86/xen/mmu_pv.c index c8dbee62ec2ab..51f8b657ec8a7 100644 --- a/arch/x86/xen/mmu_pv.c +++ b/arch/x86/xen/mmu_pv.c @@ -842,6 +842,7 @@ void xen_mm_pin_all(void) { struct page *page; + spin_lock(&init_mm.page_table_lock); spin_lock(&pgd_lock); list_for_each_entry(page, &pgd_list, lru) { @@ -852,6 +853,7 @@ void xen_mm_pin_all(void) } spin_unlock(&pgd_lock); + spin_unlock(&init_mm.page_table_lock); } static int __init xen_mark_pinned(struct mm_struct *mm, struct page *page, @@ -961,6 +963,7 @@ void xen_mm_unpin_all(void) { struct page *page; + spin_lock(&init_mm.page_table_lock); spin_lock(&pgd_lock); list_for_each_entry(page, &pgd_list, lru) { @@ -972,6 +975,7 @@ void xen_mm_unpin_all(void) } spin_unlock(&pgd_lock); + spin_unlock(&init_mm.page_table_lock); } static void xen_activate_mm(struct mm_struct *prev, struct mm_struct *next) -- 2.39.5

4 months, 3 weeks

1
0
0 0

[PATCH AUTOSEL 5.10] Grab mm lock before grabbing pt lock

by Sasha Levin

From: Maksym Planeta <maksym(a)exostellar.io> [ Upstream commit 6d002348789bc16e9203e9818b7a3688787e3b29 ] Function xen_pin_page calls xen_pte_lock, which in turn grab page table lock (ptlock). When locking, xen_pte_lock expect mm->page_table_lock to be held before grabbing ptlock, but this does not happen when pinning is caused by xen_mm_pin_all. This commit addresses lockdep warning below, which shows up when suspending a Xen VM. [ 3680.658422] Freezing user space processes [ 3680.660156] Freezing user space processes completed (elapsed 0.001 seconds) [ 3680.660182] OOM killer disabled. [ 3680.660192] Freezing remaining freezable tasks [ 3680.661485] Freezing remaining freezable tasks completed (elapsed 0.001 seconds) [ 3680.685254] [ 3680.685265] ================================== [ 3680.685269] WARNING: Nested lock was not taken [ 3680.685274] 6.12.0+ #16 Tainted: G W [ 3680.685279] ---------------------------------- [ 3680.685283] migration/0/19 is trying to lock: [ 3680.685288] ffff88800bac33c0 (ptlock_ptr(ptdesc)#2){+.+.}-{3:3}, at: xen_pin_page+0x175/0x1d0 [ 3680.685303] [ 3680.685303] but this task is not holding: [ 3680.685308] init_mm.page_table_lock [ 3680.685311] [ 3680.685311] stack backtrace: [ 3680.685316] CPU: 0 UID: 0 PID: 19 Comm: migration/0 Tainted: G W 6.12.0+ #16 [ 3680.685324] Tainted: [W]=WARN [ 3680.685328] Stopper: multi_cpu_stop+0x0/0x120 <- __stop_cpus.constprop.0+0x8c/0xd0 [ 3680.685339] Call Trace: [ 3680.685344] <TASK> [ 3680.685347] dump_stack_lvl+0x77/0xb0 [ 3680.685356] __lock_acquire+0x917/0x2310 [ 3680.685364] lock_acquire+0xce/0x2c0 [ 3680.685369] ? xen_pin_page+0x175/0x1d0 [ 3680.685373] _raw_spin_lock_nest_lock+0x2f/0x70 [ 3680.685381] ? xen_pin_page+0x175/0x1d0 [ 3680.685386] xen_pin_page+0x175/0x1d0 [ 3680.685390] ? __pfx_xen_pin_page+0x10/0x10 [ 3680.685394] __xen_pgd_walk+0x233/0x2c0 [ 3680.685401] ? stop_one_cpu+0x91/0x100 [ 3680.685405] __xen_pgd_pin+0x5d/0x250 [ 3680.685410] xen_mm_pin_all+0x70/0xa0 [ 3680.685415] xen_pv_pre_suspend+0xf/0x280 [ 3680.685420] xen_suspend+0x57/0x1a0 [ 3680.685428] multi_cpu_stop+0x6b/0x120 [ 3680.685432] ? update_cpumasks_hier+0x7c/0xa60 [ 3680.685439] ? __pfx_multi_cpu_stop+0x10/0x10 [ 3680.685443] cpu_stopper_thread+0x8c/0x140 [ 3680.685448] ? smpboot_thread_fn+0x20/0x1f0 [ 3680.685454] ? __pfx_smpboot_thread_fn+0x10/0x10 [ 3680.685458] smpboot_thread_fn+0xed/0x1f0 [ 3680.685462] kthread+0xde/0x110 [ 3680.685467] ? __pfx_kthread+0x10/0x10 [ 3680.685471] ret_from_fork+0x2f/0x50 [ 3680.685478] ? __pfx_kthread+0x10/0x10 [ 3680.685482] ret_from_fork_asm+0x1a/0x30 [ 3680.685489] </TASK> [ 3680.685491] [ 3680.685491] other info that might help us debug this: [ 3680.685497] 1 lock held by migration/0/19: [ 3680.685500] #0: ffffffff8284df38 (pgd_lock){+.+.}-{3:3}, at: xen_mm_pin_all+0x14/0xa0 [ 3680.685512] [ 3680.685512] stack backtrace: [ 3680.685518] CPU: 0 UID: 0 PID: 19 Comm: migration/0 Tainted: G W 6.12.0+ #16 [ 3680.685528] Tainted: [W]=WARN [ 3680.685531] Stopper: multi_cpu_stop+0x0/0x120 <- __stop_cpus.constprop.0+0x8c/0xd0 [ 3680.685538] Call Trace: [ 3680.685541] <TASK> [ 3680.685544] dump_stack_lvl+0x77/0xb0 [ 3680.685549] __lock_acquire+0x93c/0x2310 [ 3680.685554] lock_acquire+0xce/0x2c0 [ 3680.685558] ? xen_pin_page+0x175/0x1d0 [ 3680.685562] _raw_spin_lock_nest_lock+0x2f/0x70 [ 3680.685568] ? xen_pin_page+0x175/0x1d0 [ 3680.685572] xen_pin_page+0x175/0x1d0 [ 3680.685578] ? __pfx_xen_pin_page+0x10/0x10 [ 3680.685582] __xen_pgd_walk+0x233/0x2c0 [ 3680.685588] ? stop_one_cpu+0x91/0x100 [ 3680.685592] __xen_pgd_pin+0x5d/0x250 [ 3680.685596] xen_mm_pin_all+0x70/0xa0 [ 3680.685600] xen_pv_pre_suspend+0xf/0x280 [ 3680.685607] xen_suspend+0x57/0x1a0 [ 3680.685611] multi_cpu_stop+0x6b/0x120 [ 3680.685615] ? update_cpumasks_hier+0x7c/0xa60 [ 3680.685620] ? __pfx_multi_cpu_stop+0x10/0x10 [ 3680.685625] cpu_stopper_thread+0x8c/0x140 [ 3680.685629] ? smpboot_thread_fn+0x20/0x1f0 [ 3680.685634] ? __pfx_smpboot_thread_fn+0x10/0x10 [ 3680.685638] smpboot_thread_fn+0xed/0x1f0 [ 3680.685642] kthread+0xde/0x110 [ 3680.685645] ? __pfx_kthread+0x10/0x10 [ 3680.685649] ret_from_fork+0x2f/0x50 [ 3680.685654] ? __pfx_kthread+0x10/0x10 [ 3680.685657] ret_from_fork_asm+0x1a/0x30 [ 3680.685662] </TASK> [ 3680.685267] xen:grant_table: Grant tables using version 1 layout [ 3680.685921] OOM killer enabled. [ 3680.685934] Restarting tasks ... done. Signed-off-by: Maksym Planeta <maksym(a)exostellar.io> Reviewed-by: Juergen Gross <jgross(a)suse.com> Message-ID: <20241204103516.3309112-1-maksym(a)exostellar.io> Signed-off-by: Juergen Gross <jgross(a)suse.com> Signed-off-by: Sasha Levin <sashal(a)kernel.org> --- arch/x86/xen/mmu_pv.c | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/arch/x86/xen/mmu_pv.c b/arch/x86/xen/mmu_pv.c index cf2ade864c302..2fef8ad3f1df9 100644 --- a/arch/x86/xen/mmu_pv.c +++ b/arch/x86/xen/mmu_pv.c @@ -762,6 +762,7 @@ void xen_mm_pin_all(void) { struct page *page; + spin_lock(&init_mm.page_table_lock); spin_lock(&pgd_lock); list_for_each_entry(page, &pgd_list, lru) { @@ -772,6 +773,7 @@ void xen_mm_pin_all(void) } spin_unlock(&pgd_lock); + spin_unlock(&init_mm.page_table_lock); } static void __init xen_mark_pinned(struct mm_struct *mm, struct page *page, @@ -866,6 +868,7 @@ void xen_mm_unpin_all(void) { struct page *page; + spin_lock(&init_mm.page_table_lock); spin_lock(&pgd_lock); list_for_each_entry(page, &pgd_list, lru) { @@ -877,6 +880,7 @@ void xen_mm_unpin_all(void) } spin_unlock(&pgd_lock); + spin_unlock(&init_mm.page_table_lock); } static void xen_activate_mm(struct mm_struct *prev, struct mm_struct *next) -- 2.39.5

4 months, 3 weeks

1
0
0 0

[PATCH AUTOSEL 5.15] Grab mm lock before grabbing pt lock

by Sasha Levin

From: Maksym Planeta <maksym(a)exostellar.io> [ Upstream commit 6d002348789bc16e9203e9818b7a3688787e3b29 ] Function xen_pin_page calls xen_pte_lock, which in turn grab page table lock (ptlock). When locking, xen_pte_lock expect mm->page_table_lock to be held before grabbing ptlock, but this does not happen when pinning is caused by xen_mm_pin_all. This commit addresses lockdep warning below, which shows up when suspending a Xen VM. [ 3680.658422] Freezing user space processes [ 3680.660156] Freezing user space processes completed (elapsed 0.001 seconds) [ 3680.660182] OOM killer disabled. [ 3680.660192] Freezing remaining freezable tasks [ 3680.661485] Freezing remaining freezable tasks completed (elapsed 0.001 seconds) [ 3680.685254] [ 3680.685265] ================================== [ 3680.685269] WARNING: Nested lock was not taken [ 3680.685274] 6.12.0+ #16 Tainted: G W [ 3680.685279] ---------------------------------- [ 3680.685283] migration/0/19 is trying to lock: [ 3680.685288] ffff88800bac33c0 (ptlock_ptr(ptdesc)#2){+.+.}-{3:3}, at: xen_pin_page+0x175/0x1d0 [ 3680.685303] [ 3680.685303] but this task is not holding: [ 3680.685308] init_mm.page_table_lock [ 3680.685311] [ 3680.685311] stack backtrace: [ 3680.685316] CPU: 0 UID: 0 PID: 19 Comm: migration/0 Tainted: G W 6.12.0+ #16 [ 3680.685324] Tainted: [W]=WARN [ 3680.685328] Stopper: multi_cpu_stop+0x0/0x120 <- __stop_cpus.constprop.0+0x8c/0xd0 [ 3680.685339] Call Trace: [ 3680.685344] <TASK> [ 3680.685347] dump_stack_lvl+0x77/0xb0 [ 3680.685356] __lock_acquire+0x917/0x2310 [ 3680.685364] lock_acquire+0xce/0x2c0 [ 3680.685369] ? xen_pin_page+0x175/0x1d0 [ 3680.685373] _raw_spin_lock_nest_lock+0x2f/0x70 [ 3680.685381] ? xen_pin_page+0x175/0x1d0 [ 3680.685386] xen_pin_page+0x175/0x1d0 [ 3680.685390] ? __pfx_xen_pin_page+0x10/0x10 [ 3680.685394] __xen_pgd_walk+0x233/0x2c0 [ 3680.685401] ? stop_one_cpu+0x91/0x100 [ 3680.685405] __xen_pgd_pin+0x5d/0x250 [ 3680.685410] xen_mm_pin_all+0x70/0xa0 [ 3680.685415] xen_pv_pre_suspend+0xf/0x280 [ 3680.685420] xen_suspend+0x57/0x1a0 [ 3680.685428] multi_cpu_stop+0x6b/0x120 [ 3680.685432] ? update_cpumasks_hier+0x7c/0xa60 [ 3680.685439] ? __pfx_multi_cpu_stop+0x10/0x10 [ 3680.685443] cpu_stopper_thread+0x8c/0x140 [ 3680.685448] ? smpboot_thread_fn+0x20/0x1f0 [ 3680.685454] ? __pfx_smpboot_thread_fn+0x10/0x10 [ 3680.685458] smpboot_thread_fn+0xed/0x1f0 [ 3680.685462] kthread+0xde/0x110 [ 3680.685467] ? __pfx_kthread+0x10/0x10 [ 3680.685471] ret_from_fork+0x2f/0x50 [ 3680.685478] ? __pfx_kthread+0x10/0x10 [ 3680.685482] ret_from_fork_asm+0x1a/0x30 [ 3680.685489] </TASK> [ 3680.685491] [ 3680.685491] other info that might help us debug this: [ 3680.685497] 1 lock held by migration/0/19: [ 3680.685500] #0: ffffffff8284df38 (pgd_lock){+.+.}-{3:3}, at: xen_mm_pin_all+0x14/0xa0 [ 3680.685512] [ 3680.685512] stack backtrace: [ 3680.685518] CPU: 0 UID: 0 PID: 19 Comm: migration/0 Tainted: G W 6.12.0+ #16 [ 3680.685528] Tainted: [W]=WARN [ 3680.685531] Stopper: multi_cpu_stop+0x0/0x120 <- __stop_cpus.constprop.0+0x8c/0xd0 [ 3680.685538] Call Trace: [ 3680.685541] <TASK> [ 3680.685544] dump_stack_lvl+0x77/0xb0 [ 3680.685549] __lock_acquire+0x93c/0x2310 [ 3680.685554] lock_acquire+0xce/0x2c0 [ 3680.685558] ? xen_pin_page+0x175/0x1d0 [ 3680.685562] _raw_spin_lock_nest_lock+0x2f/0x70 [ 3680.685568] ? xen_pin_page+0x175/0x1d0 [ 3680.685572] xen_pin_page+0x175/0x1d0 [ 3680.685578] ? __pfx_xen_pin_page+0x10/0x10 [ 3680.685582] __xen_pgd_walk+0x233/0x2c0 [ 3680.685588] ? stop_one_cpu+0x91/0x100 [ 3680.685592] __xen_pgd_pin+0x5d/0x250 [ 3680.685596] xen_mm_pin_all+0x70/0xa0 [ 3680.685600] xen_pv_pre_suspend+0xf/0x280 [ 3680.685607] xen_suspend+0x57/0x1a0 [ 3680.685611] multi_cpu_stop+0x6b/0x120 [ 3680.685615] ? update_cpumasks_hier+0x7c/0xa60 [ 3680.685620] ? __pfx_multi_cpu_stop+0x10/0x10 [ 3680.685625] cpu_stopper_thread+0x8c/0x140 [ 3680.685629] ? smpboot_thread_fn+0x20/0x1f0 [ 3680.685634] ? __pfx_smpboot_thread_fn+0x10/0x10 [ 3680.685638] smpboot_thread_fn+0xed/0x1f0 [ 3680.685642] kthread+0xde/0x110 [ 3680.685645] ? __pfx_kthread+0x10/0x10 [ 3680.685649] ret_from_fork+0x2f/0x50 [ 3680.685654] ? __pfx_kthread+0x10/0x10 [ 3680.685657] ret_from_fork_asm+0x1a/0x30 [ 3680.685662] </TASK> [ 3680.685267] xen:grant_table: Grant tables using version 1 layout [ 3680.685921] OOM killer enabled. [ 3680.685934] Restarting tasks ... done. Signed-off-by: Maksym Planeta <maksym(a)exostellar.io> Reviewed-by: Juergen Gross <jgross(a)suse.com> Message-ID: <20241204103516.3309112-1-maksym(a)exostellar.io> Signed-off-by: Juergen Gross <jgross(a)suse.com> Signed-off-by: Sasha Levin <sashal(a)kernel.org> --- arch/x86/xen/mmu_pv.c | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/arch/x86/xen/mmu_pv.c b/arch/x86/xen/mmu_pv.c index 3359c23573c50..f737e3af3edcc 100644 --- a/arch/x86/xen/mmu_pv.c +++ b/arch/x86/xen/mmu_pv.c @@ -762,6 +762,7 @@ void xen_mm_pin_all(void) { struct page *page; + spin_lock(&init_mm.page_table_lock); spin_lock(&pgd_lock); list_for_each_entry(page, &pgd_list, lru) { @@ -772,6 +773,7 @@ void xen_mm_pin_all(void) } spin_unlock(&pgd_lock); + spin_unlock(&init_mm.page_table_lock); } static void __init xen_mark_pinned(struct mm_struct *mm, struct page *page, @@ -866,6 +868,7 @@ void xen_mm_unpin_all(void) { struct page *page; + spin_lock(&init_mm.page_table_lock); spin_lock(&pgd_lock); list_for_each_entry(page, &pgd_list, lru) { @@ -877,6 +880,7 @@ void xen_mm_unpin_all(void) } spin_unlock(&pgd_lock); + spin_unlock(&init_mm.page_table_lock); } static void xen_activate_mm(struct mm_struct *prev, struct mm_struct *next) -- 2.39.5

4 months, 3 weeks

1
0
0 0

2025

2024

2023

2022

2021

2020

2019

2018

2017

Linux-stable-mirror February 2025