ARM: mm: fix race updating mm->context.id on ASID rollover
PATCH 1/2 ARM: mm: fix race updating mm->context.id on ASID rollover
If a thread triggers an ASID rollover, other threads of the same process must be made to wait until the mm->context.id for the shared mm_struct has been updated to new generation and associated book-keeping (e.g. TLB invalidation) has ben performed.
However, there is a *tiny* window where both mm->context.id and the relevant active_asids entry are updated to the new generation, but the TLB flush has not been performed, which could allow another thread to return to userspace with a dirty TLB, potentially leading to data corruption. In reality this will never occur because one CPU would need to perform a context-switch in the time it takes another to do a couple of atomic test/set operations but we should plug the race anyway.
This patch moves the active_asids update until after the potential TLB flush on context-switch.
--- arch/arm/mm/context.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
arch/arm/mm/context.c +3 -3
@@ -207,11 +207,11 @@ void check_and_switch_context(struct mm_struct *mm, struct task_struct *tsk)if ((mm->context.id ^atomic64_read (&asid_generation)) >>ASID_BITS )new_context (mm, cpu); -atomic64_set (&per_cpu (active_asids, cpu), mm->context.id); -cpumask_set_cpu (cpu,mm_cpumask (mm)); -if (cpumask_test_and_clear_cpu (cpu, &tlb_flush_pending))local_flush_tlb_all (); + +atomic64_set (&per_cpu (active_asids, cpu), mm->context.id); +cpumask_set_cpu (cpu,mm_cpumask (mm));raw_spin_unlock_irqrestore (&cpu_asid_lock, flags); switch_mm_fastpath:
Signature · 1 line
1.8.0
PATCH 2/2 ARM: mm: make mm->context.id an atomic64_t variable
mm->context.id is updated under asid_lock when a new ASID is allocated to an mm_struct. However, it is also read without the lock when a task is being scheduled and checking whether or not the current ASID generation is up-to-date.
If two threads of the same process are being scheduled in parallel and the bottom bits of the generation in their mm->context.id match the current generation (that is, the mm_struct has not been used for ~2^24 rollovers) then the non-atomic, lockless access to mm->context.id may yield the incorrect ASID.
This patch fixes this issue by making mm->context.id and atomic64_t, ensuring that the generation is always read consistently. For code that only requires access to the ASID bits (e.g. TLB flushing by mm), then the value is accessed directly, which GCC converts to an ldrb.
--- arch/arm/include/asm/mmu.h | 8 ++++---- arch/arm/include/asm/mmu_context.h | 2 +- arch/arm/kernel/asm-offsets.c | 2 +- arch/arm/mm/context.c | 21 +++++++++++++-------- 4 files changed, 19 insertions(+), 14 deletions(-)
arch/arm/include/asm/mmu.h +4 -4
@@ -5,15 +5,15 @@typedef struct {#ifdef CONFIG_CPU_HAS_ASID -u64 id; +atomic64_t id;#endif -unsigned int vmalloc_seq; +unsigned int vmalloc_seq; }mm_context_t ;#ifdef CONFIG_CPU_HAS_ASID #define ASID_BITS 8#define ASID_MASK ((~0ULL) << ASID_BITS) -#define ASID (mm) ((mm)->context.id & ~ASID_MASK) +#define ASID (mm) ((mm)->context.id.counter & ~ASID_MASK)#else #define ASID (mm) (0)#endif
@@ -26,7 +26,7 @@ typedef struct {* modified for 2.6 by Hyok S. Choi <[email protected]> */ typedef struct { -unsigned long end_brk; +unsigned long end_brk; }mm_context_t ;#endif
arch/arm/include/asm/mmu_context.h +1 -1
@@ -25,7 +25,7 @@ void __check_vmalloc_seq(struct mm_struct *mm);#ifdef CONFIG_CPU_HAS_ASID void check_and_switch_context (struct mm_struct *mm,struct task_struct *tsk); -#define init_new_context (tsk,mm) ({ mm->context.id = 0; }) +#define init_new_context (tsk,mm) ({ atomic64_set(&mm->context.id, 0); 0; })#else /* !CONFIG_CPU_HAS_ASID */
arch/arm/kernel/asm-offsets.c +1 -1
@@ -110,7 +110,7 @@ int main(void)BLANK ();#endif #ifdef CONFIG_CPU_HAS_ASID - DEFINE(MM_CONTEXT_ID, offsetof(struct mm_struct, context.id)); + DEFINE(MM_CONTEXT_ID, offsetof(struct mm_struct, context.id.counter));BLANK ();#endif DEFINE (VMA_VM_MM , offsetof(struct vm_area_struct , vm_mm));
arch/arm/mm/context.c +13 -8
@@ -152,9 +152,9 @@ static int is_reserved_asid(u64 asid)return 0 ; } -static void new_context (struct mm_struct *mm,unsigned int cpu) +static u64 new_context (struct mm_struct *mm,unsigned int cpu) { -u64 asid = mm->context.id; +u64 asid =atomic64_read (&mm->context.id);u64 generation =atomic64_read (&asid_generation);if (asid !=0 &&is_reserved_asid (asid)) {
@@ -181,13 +181,14 @@ static void new_context(struct mm_struct *mm, unsigned int cpu)cpumask_clear (mm_cpumask (mm)); } - mm->context.id = asid; +return asid; }void check_and_switch_context (struct mm_struct *mm,struct task_struct *tsk) {unsigned long flags;unsigned int cpu =smp_processor_id (); +u64 asid;if (unlikely (mm->context.vmalloc_seq != init_mm.context.vmalloc_seq))__check_vmalloc_seq (mm);
@@ -198,19 +199,23 @@ void check_and_switch_context(struct mm_struct *mm, struct task_struct *tsk)*/ cpu_set_reserved_ttbr0 (); -if (!((mm->context.id ^atomic64_read (&asid_generation)) >>ASID_BITS ) - &&atomic64_xchg (&per_cpu (active_asids, cpu), mm->context.id)) + asid =atomic64_read (&mm->context.id); +if (!((asid ^atomic64_read (&asid_generation)) >>ASID_BITS ) + &&atomic64_xchg (&per_cpu (active_asids, cpu), asid)) goto switch_mm_fastpath;raw_spin_lock_irqsave (&cpu_asid_lock, flags);/* Check that our ASID belongs to the current generation. */ -if ((mm->context.id ^atomic64_read (&asid_generation)) >>ASID_BITS ) -new_context (mm, cpu); + asid =atomic64_read (&mm->context.id); +if ((asid ^atomic64_read (&asid_generation)) >>ASID_BITS ) { + asid =new_context (mm, cpu); +atomic64_set (&mm->context.id, asid); + }if (cpumask_test_and_clear_cpu (cpu, &tlb_flush_pending))local_flush_tlb_all (); -atomic64_set (&per_cpu (active_asids, cpu), mm->context.id); +atomic64_set (&per_cpu (active_asids, cpu), asid);cpumask_set_cpu (cpu,mm_cpumask (mm));raw_spin_unlock_irqrestore (&cpu_asid_lock, flags);
Signature · 1 line
1.8.0
PATCH 1/2 ARM: mm: fix race updating mm->context.id on ASID rollover
In reply to Will DeaconQuoted 13 lines · Will Deacon
If a thread triggers an ASID rollover, other threads of the same process must be made to wait until the
mm->context.idfor the sharedmm_structhas been updated to new generation and associated book-keeping (e.g. TLB invalidation) has ben performed.However, there is a *tiny* window where both
mm->context.idand the relevantactive_asidsentry are updated to the new generation, but the TLB flush has not been performed, which could allow another thread to return to userspace with a dirty TLB, potentially leading to data corruption. In reality this will never occur because one CPU would need to perform a context-switch in the time it takes another to do a couple of atomic test/set operations but we should plug the race anyway.
Quoted 5 lines
This patch moves the
active_asidsupdate until after the potential TLB flush on context-switch.Cc <[email protected]>Signed-off-by Will Deacon <[email protected]>
PATCH 2/2 ARM: mm: make mm->context.id an atomic64_t variable
In reply to Will DeaconQuoted 11 lines · Will Deacon
mm->context.idis updated underasid_lockwhen a new ASID is allocated to anmm_struct. However, it is also read without the lock when a task is being scheduled and checking whether or not the current ASID generation is up-to-date.If two threads of the same process are being scheduled in parallel and the bottom bits of the generation in their
mm->context.idmatch the current generation (that is, themm_structhas not been used for ~2^24 rollovers) then the non-atomic, lockless access tomm->context.idmay yield the incorrect ASID.
Quoted 7 lines
This patch fixes this issue by making
mm->context.idandatomic64_t, ensuring that the generation is always read consistently. For code that only requires access to the ASID bits (e.g. TLB flushing by mm), then the value is accessed directly, which GCC converts to an ldrb.Cc <[email protected]>Signed-off-by Will Deacon <[email protected]>