mm: Unconditional per-VMA locks and cleanups
Subsystems (25)
Suren Baghdasaryan
maintainerv7 cover mm: Unconditional per-VMA locks and cleanups
v2 version of this patchset [1] was written by Dave Hansen and per his request, I'm taking over this series.
tl;dr: Make per-VMA locks available in all configs. Simplify some of the per-VMA lock users now that they can rely on them being always available.
Binder and networking folks: Your code is the target of the cleanups. I'm cc'ing you now on v2 because there's emerging consensus on the mm side that the approach here is sane. I'm not quite sure how this pile would get merged, but ack/review tags would be appreciated if this looks good to you.
Longer version:
When working on some x86 shadow stack code, it was a real pain to avoid causing recursive locking problems with mmap_lock. One way to avoid those was to avoid mmap_lock and use per-VMA locks instead. They are great, but they are not available in all configs which makes them unusable in generic code, or if you want to completely avoid mmap_lock.
Make per-VMA locks available in all configs. Right now, they are only available on select architectures when SMP and MMU are enabled. But all of the primitives that per-VMA locks are built on (RCU, maple trees, refcounts) work just fine without SMP or MMU.
The only real downside is that making VMAs a wee bit bigger on !MMU and !SMP builds.
The upside is much cleaner code, lower complexity and less #ifdeffery.
Clean up a binder VMA locking site now that it can rely on per-VMA locks.
Building on top of universally-available per-VMA locks, introduce a new helper. Since the new API does not require callers to have a fallback to mmap_lock, it's much easier to use. Callers can potentially replace this very common kernel idiom:
mmap_read_lock(mm); vma = vma_lookup() // fiddle with vma mmap_read_unlock(mm);
with:
vma = vma_start_read_unlocked(mm, address); // fiddle with vma vma_end_read(vma);
Which avoids mmap_lock entirely in the fast path.
Use that new API for another binder site and one in the TCP code.
Changes since v6 [2]:
- Rebased over mm-unstable
Patch 1:
- Added Reviewed-by, per Lorenzo Stoakes Patch 2:
- Added Acked-by, per Carlos Llamas
Patch 4:
- Added INVARIANT comment, per Sashiko and Alice Ryhl
Applies cleanly over mm-unstable
[1] https://lore.kernel.org/all/[email protected]/
[2] https://lore.kernel.org/all/[email protected]/
Dave Hansen (5):
mm: Make per-VMA locks available universally
binder: Make shrinker rely solely on per-VMA lock
mm: Add RCU-based VMA lookup helper that waits for writers
binder: Remove mmap_lock fallback
tcp: Remove mmap_lock fallback path
arch/arm/Kconfig | 1 - arch/arm64/Kconfig | 1 - arch/loongarch/Kconfig | 1 - arch/powerpc/platforms/powernv/Kconfig | 1 - arch/powerpc/platforms/pseries/Kconfig | 1 - arch/riscv/Kconfig | 1 - arch/s390/Kconfig | 1 - arch/x86/Kconfig | 2 - drivers/android/binder/page_range.rs | 19 +----- drivers/android/binder_alloc.c | 63 +++++++---------- fs/proc/internal.h | 2 - fs/proc/task_mmu.c | 93 ------------------------- include/linux/mm.h | 12 ---- include/linux/mm_types.h | 8 +-- include/linux/mmap_lock.h | 94 +++++++++++--------------- kernel/bpf/stackmap.c | 17 ++--- kernel/bpf/task_iter.c | 2 +- kernel/fork.c | 2 - mm/Kconfig | 12 ---- mm/Kconfig.debug | 1 - mm/debug.c | 4 -- mm/init-mm.c | 2 - mm/memory.c | 2 - mm/mmap_lock.c | 61 ++++++++++------- mm/pagewalk.c | 2 - mm/rmap.c | 2 - mm/userfaultfd.c | 61 ++--------------- net/ipv4/tcp.c | 31 +++------ rust/kernel/mm.rs | 58 ++++++++++------ tools/testing/vma/include/dup.h | 5 +- tools/testing/vma/vma_internal.h | 1 - 31 files changed, 167 insertions(+), 396 deletions(-)
base-commit: 42d64d4fef83a241c919c8693fdf0a21b2cb6061
Signature · 1 line
2.55.0.966.g6673acef38-goog
Suren Baghdasaryan
maintainerPATCH v7 1/5 mm: Make per-VMA locks available universally
From: Dave Hansen <[email protected]>
The per-VMA locks have been around for several years. They've had some bugs worked out of them and have seen quite wide use. However, they are still only available when architectures explicitly enable them. Remove the conditional compilation around the per-VMA locks, making them available on all architectures and configs.
The approach up to now seemed to be to add ARCH_SUPPORTS_PER_VMA_LOCK when the architecture started using per-VMA locks in the fault handler. But, contrary to the naming, the Kconfig option does not really indicate whether the architecture supports per-VMA locks or not. It is more of a marker for whether the architecture is likely to benefit from per-VMA locks.
To me, the most important thing side-effect of universal availability is letting per-VMA locks be used in SMP=n configs. This lets us use per-VMA locking in all x86 code without fallbacks.
Overall, this just generally makes the kernel simpler. Just look at the diffstat. It also opens the door to users that want to use the per-VMA locks in common code. Doing *that* brings additional simplifications.
The downside of this is adding some fields to vm_area_struct and mm_struct. There are likely ways to optimize this, especially for things like SMP=n configs. For now, do the simplest thing: use the same implementation everywhere.
== Considerations for NOMMU config ==
NOMMU systems do not write-lock VMAs, therefore read-locking a VMA would always succeed unless VMA is detached. Therefore for NOMMU config we make vma_mark_attached() a NOOP, which keeps VMAs always in detached state. This causes VMA read-locking to always fail and the caller falls back to locking mmap_lock.
The following functions will have a different implementation in NOMMU config:
- vma_mark_attached(), vma_mark_detached() are made NOOPs, keeping VMAs always in a detached state and preventing assertions and refcount underflows;
- vma_start_write(), vma_start_write_killable() are made NOOPs to avoid warnings in __vma_start_write() due to VMAs being detached. These functions are not used in NOMMU code but __vma_start_write() is an exported function, therefore might be used by drivers.
- vma_assert_attached() is made NOOP because it's reachable from NOMMU code via split_vma()->vma_iter_store_new()->vma_iter_store_overwrite();
- vma_assert_write_locked() is asserting vma->vm_mm is write-locked, as was done before this change;
- vma_assert_locked() is asserting vma->vm_mm is locked, as was done before this change;
The following functions work for both MMU and NOMMU configs:
- vma_lock_init() performs the same initialization as for MMU config;
- mm_lock_seqcount_init(), mm_lock_seqcount_begin(), mm_lock_seqcount_end() are called from mmap_write_{lock|unlock} and update mm_lock_seq correctly.
- mmap_lock_speculate_try_begin(), mmap_lock_speculate_retry() work as is because mm_lock_seq is updated correctly;
- vma_start_read(), vma_start_read_locked() will always fail because VMAs are always detached;
- vma_end_read() will never be called because vma_start_read() never succeeds;
- vma_is_attached() always return false because VMAs are always detached;
- vma_assert_detached() will never trigger because VMAs are never attached;
- vma_start_read_locked() always return false because VMAs are always detached;
- lock_vma_under_rcu() will be safe as the attempted read lock will bail;
Changes in the following files are not affecting NOMMU config:
task_mmu.c - not compiled when CONFIG_MMU=n;pagewalk.c - not compiled when CONFIG_MMU=n;userfaultfd.c - not compiled when CONFIG_MMU=n (CONFIG_USERFAULTFD depends on CONFIG_MMU);
The following changes in the BPF code are made to keep NOMMU config working like before:
stack_map_lock_vma() - keeps mmap_lock in NOMMU config; bpf_iter_task_vma_new() - bails out in NOMMU config;
--- arch/arm/Kconfig | 1 - arch/arm64/Kconfig | 1 - arch/loongarch/Kconfig | 1 - arch/powerpc/platforms/powernv/Kconfig | 1 - arch/powerpc/platforms/pseries/Kconfig | 1 - arch/riscv/Kconfig | 1 - arch/s390/Kconfig | 1 - arch/x86/Kconfig | 2 - fs/proc/internal.h | 2 - fs/proc/task_mmu.c | 93 -------------------------- include/linux/mm.h | 12 ---- include/linux/mm_types.h | 8 +-- include/linux/mmap_lock.h | 75 +++++++-------------- kernel/bpf/stackmap.c | 17 ++--- kernel/bpf/task_iter.c | 2 +- kernel/fork.c | 2 - mm/Kconfig | 12 ---- mm/Kconfig.debug | 1 - mm/debug.c | 4 -- mm/init-mm.c | 2 - mm/memory.c | 2 - mm/mmap_lock.c | 26 +------ mm/pagewalk.c | 2 - mm/rmap.c | 2 - mm/userfaultfd.c | 55 --------------- rust/kernel/mm.rs | 32 +++------ tools/testing/vma/include/dup.h | 5 +- tools/testing/vma/vma_internal.h | 1 - 28 files changed, 48 insertions(+), 316 deletions(-)
arch/arm/Kconfig +0 -1
@@ -42,7 +42,6 @@ config ARM select ARCH_SUPPORTS_ATOMIC_RMW select ARCH_SUPPORTS_CFI select ARCH_SUPPORTS_HUGETLBFS if ARM_LPAE - select ARCH_SUPPORTS_PER_VMA_LOCK select ARCH_SUPPORTS_RT select ARCH_USE_BUILTIN_BSWAP select ARCH_USE_CMPXCHG_LOCKREF
arch/arm64/Kconfig +0 -1
@@ -81,7 +81,6 @@ config ARM64 select ARCH_HAS_PTE_PROTNONE select ARCH_SUPPORTS_NUMA_BALANCING select ARCH_SUPPORTS_PAGE_TABLE_CHECK - select ARCH_SUPPORTS_PER_VMA_LOCK select ARCH_SUPPORTS_HUGE_PFNMAP if TRANSPARENT_HUGEPAGE select ARCH_SUPPORTS_RT select ARCH_SUPPORTS_SCHED_SMT
arch/loongarch/Kconfig +0 -1
@@ -69,7 +69,6 @@ config LOONGARCH select ARCH_SUPPORTS_MSEAL_SYSTEM_MAPPINGS select ARCH_HAS_PTE_PROTNONE if 64BIT select ARCH_SUPPORTS_NUMA_BALANCING if NUMA - select ARCH_SUPPORTS_PER_VMA_LOCK select ARCH_SUPPORTS_RT select ARCH_SUPPORTS_SCHED_SMT if SMP select ARCH_SUPPORTS_SCHED_MC if SMP
arch/powerpc/platforms/powernv/Kconfig +0 -1
@@ -17,7 +17,6 @@ config PPC_POWERNV select PPC_DOORBELL select MMU_NOTIFIER select FORCE_SMP - select ARCH_SUPPORTS_PER_VMA_LOCK select PPC_RADIX_BROADCAST_TLBIE if PPC_RADIX_MMU default y
arch/powerpc/platforms/pseries/Kconfig +0 -1
@@ -23,7 +23,6 @@ config PPC_PSERIES select HOTPLUG_CPU select FORCE_SMP select SWIOTLB - select ARCH_SUPPORTS_PER_VMA_LOCK select PPC_RADIX_BROADCAST_TLBIE if PPC_RADIX_MMU default y
arch/riscv/Kconfig +0 -1
@@ -72,7 +72,6 @@ config RISCV select ARCH_SUPPORTS_LTO_CLANG_THIN select ARCH_SUPPORTS_MSEAL_SYSTEM_MAPPINGS if 64BIT && MMU select ARCH_SUPPORTS_PAGE_TABLE_CHECK if MMU - select ARCH_SUPPORTS_PER_VMA_LOCK if MMU select ARCH_HAS_PTE_PROTNONE if MMU select ARCH_SUPPORTS_RT select ARCH_SUPPORTS_SHADOW_CALL_STACK if HAVE_SHADOW_CALL_STACK
arch/s390/Kconfig +0 -1
@@ -156,7 +156,6 @@ config S390 select ARCH_HAS_PTE_PROTNONE select ARCH_SUPPORTS_NUMA_BALANCING select ARCH_SUPPORTS_PAGE_TABLE_CHECK - select ARCH_SUPPORTS_PER_VMA_LOCK select ARCH_USES_CFI_GENERIC_LLVM_PASS if CC_IS_CLANG select ARCH_USE_BUILTIN_BSWAP select ARCH_USE_CMPXCHG_LOCKREF
arch/x86/Kconfig +0 -2
@@ -27,7 +27,6 @@ config X86_64 select ARCH_HAS_GIGANTIC_PAGE select ARCH_SUPPORTS_MSEAL_SYSTEM_MAPPINGS select ARCH_SUPPORTS_INT128 if CC_HAS_INT128 - select ARCH_SUPPORTS_PER_VMA_LOCK select ARCH_SUPPORTS_HUGE_PFNMAP if TRANSPARENT_HUGEPAGE select HAVE_ARCH_SOFT_DIRTY select MODULES_USE_ELF_RELA
@@ -1849,7 +1848,6 @@ config X86_USER_SHADOW_STACK bool "X86 userspace shadow stack" depends on AS_WRUSS depends on X86_64 - depends on PER_VMA_LOCK select ARCH_USES_HIGH_VMA_FLAGS select ARCH_HAS_USER_SHADOW_STACK select X86_CET
fs/proc/internal.h +0 -2
@@ -385,10 +385,8 @@ struct mem_size_stats;struct proc_maps_locking_ctx {struct mm_struct *mm; -#ifdef CONFIG_PER_VMA_LOCK bool mmap_locked;struct vm_area_struct *locked_vma; -#endif };struct proc_maps_private {
fs/proc/task_mmu.c +0 -93
141 lines not on this page · raw message
include/linux/mm.h +0 -12
@@ -928,7 +928,6 @@ static inline void vma_numab_state_free(struct vm_area_struct *vma) {}* These must be here rather than mmap_lock.h as dependent on vm_fault type, * declared in this header. */ -#ifdef CONFIG_PER_VMA_LOCK static inline void release_fault_lock (struct vm_fault *vmf) {if (vmf->flags &FAULT_FLAG_VMA_LOCK )
@@ -944,17 +943,6 @@ static inline void assert_fault_locked(const struct vm_fault *vmf)else mmap_assert_locked (vmf->vma->vm_mm); } -#else -static inline void release_fault_lock (struct vm_fault *vmf) -{ -mmap_read_unlock (vmf->vma->vm_mm); -} - -static inline void assert_fault_locked (const struct vm_fault *vmf) -{ -mmap_assert_locked (vmf->vma->vm_mm); -} -#endif /* CONFIG_PER_VMA_LOCK */ static inline bool mm_flags_test (int flag,const struct mm_struct *mm) {
include/linux/mm_types.h +2 -6
@@ -950,7 +950,6 @@ struct vm_area_struct {vma_flags_t flags; }; -#ifdef CONFIG_PER_VMA_LOCK /* * Can only be written (using WRITE_ONCE()) while holding both: * - mmap_lock (in write mode)
@@ -966,7 +965,7 @@ struct vm_area_struct {* slowpath. */ unsigned int vm_lock_seq; -#endif +/* * Low 32-bits of anonymous page offset. * See vma_start_anon_pgoff() comment for details.
@@ -1003,7 +1002,6 @@ struct vm_area_struct {#ifdef CONFIG_NUMA_BALANCING struct vma_numab_state *numab_state;/* NUMA Balancing state */ #endif -#ifdef CONFIG_PER_VMA_LOCK /* * Used to keep track of firstly, whether the VMA is attached, secondly, * if attached, how many read locks are taken, and thirdly, if the
@@ -1046,7 +1044,6 @@ struct vm_area_struct {#ifdef CONFIG_DEBUG_LOCK_ALLOC struct lockdep_map vmlock_dep_map;#endif -#endif #ifdef CONFIG_64BIT /* * High 32-bits of anonymous page offset.
@@ -1254,7 +1251,6 @@ struct mm_struct {* init_mm.mmlist, and are protected * by mmlist_lock */ -#ifdef CONFIG_PER_VMA_LOCK struct rcuwait vma_writer_wait;/* * This field has lock-like semantics, meaning it is sometimes
@@ -1274,7 +1270,7 @@ struct mm_struct {* mmap_lock. */ seqcount_t mm_lock_seq; -#endif +struct futex_mm_data futex;unsigned long hiwater_rss;/* High-watermark of RSS usage */
include/linux/mmap_lock.h +25 -50
129 lines not on this page · raw message
kernel/bpf/stackmap.c +6 -11
@@ -272,13 +272,10 @@ struct stack_map_vma_lock {/* * Acquire a stable read-side reference on the VMA covering @ip. * -* With CONFIG_PER_VMA_LOCK=y this returns a VMA with its per-VMA read -* lock held and mmap_lock dropped, so the caller may sleep. -* -* With CONFIG_PER_VMA_LOCK=n it returns a VMA with mmap_lock still -* held; the caller must snapshot any fields it needs and pin vm_file -* with get_file() before stack_map_unlock_vma() drops mmap_lock, as -* the VMA may be split, merged, or freed after that. +* On NOMMU configurations, returns with the mmap_lock held. If the MMU +* is enabled, the per-VMA lock will be held instead. The lock +* should be released with stack_map_unlock_vma() which will release the +* appropriate lock. Once the lock is released, the VMA may be freed. * * Returns NULL on failure, in which case no lock is held. */
@@ -288,7 +285,6 @@ stack_map_lock_vma(struct stack_map_vma_lock *lock, unsigned long ip)struct mm_struct *mm = lock->mm;struct vm_area_struct *vma; -/* noop under !CONFIG_PER_VMA_LOCK */ vma =lock_vma_under_rcu (mm, ip);if (vma) { lock->vma = vma;
@@ -308,21 +304,20 @@ stack_map_lock_vma(struct stack_map_vma_lock *lock, unsigned long ip)return NULL ; } -#ifdef CONFIG_PER_VMA_LOCK +#ifdef CONFIG_MMU if (!vma_start_read_locked (vma)) {mmap_read_unlock (mm);return NULL ; }mmap_read_unlock (mm);#endif - lock->vma = vma;return vma; }static void stack_map_unlock_vma (struct stack_map_vma_lock *lock) { -#ifdef CONFIG_PER_VMA_LOCK +#ifdef CONFIG_MMU vma_end_read (lock->vma);#else mmap_read_unlock (lock->mm);
kernel/bpf/task_iter.c +1 -1
@@ -869,7 +869,7 @@ __bpf_kfunc int bpf_iter_task_vma_new(struct bpf_iter_task_vma *it,BUILD_BUG_ON (sizeof (struct bpf_iter_task_vma_kern ) !=sizeof (struct bpf_iter_task_vma ));BUILD_BUG_ON (__alignof__(struct bpf_iter_task_vma_kern ) != __alignof__(struct bpf_iter_task_vma )); -if (!IS_ENABLED (CONFIG_PER_VMA_LOCK )) { +if (!IS_ENABLED (CONFIG_MMU )) { kit->data =NULL ;return -EOPNOTSUPP ; }
kernel/fork.c +0 -2
@@ -1083,9 +1083,7 @@ static void mmap_init_lock(struct mm_struct *mm) {init_rwsem (&mm->mmap_lock);mm_lock_seqcount_init (mm); -#ifdef CONFIG_PER_VMA_LOCK rcuwait_init (&mm->vma_writer_wait); -#endif }static struct mm_struct *mm_init (struct mm_struct *mm,struct task_struct *p)
mm/Kconfig +0 -12
@@ -1430,18 +1430,6 @@ config LRU_GEN_WALKS_MMU depends on LRU_GEN && ARCH_HAS_HW_PTE_YOUNG # } -config ARCH_SUPPORTS_PER_VMA_LOCK - def_bool n - -config PER_VMA_LOCK - def_bool y - depends on ARCH_SUPPORTS_PER_VMA_LOCK && MMU && SMP - help - Allow per-vma locking during page fault handling. - - This feature allows locking each virtual memory area separately when - handling page faults instead of taking mmap_lock. - config LOCK_MM_AND_FIND_VMA bool depends on !STACK_GROWSUP
mm/Kconfig.debug +0 -1
@@ -310,7 +310,6 @@ config DEBUG_KMEMLEAK_VERBOSE config PER_VMA_LOCK_STATS bool "Statistics for per-vma locks" - depends on PER_VMA_LOCK help Say Y here to enable success, retry and failure counters of page faults handled under protection of per-vma locks. When enabled, the
mm/debug.c +0 -4
@@ -157,17 +157,13 @@ void dump_vma(const struct vm_area_struct *vma)pr_emerg ("vma %px start %px end %px mm %px\n" "prot %lx anon_vma %px vm_ops %px\n" "pgoff %lx file %px private_data %px\n" -#ifdef CONFIG_PER_VMA_LOCK"refcnt %x\n" -#endif"flags: %#lx(%pGv)\n" , vma, (void *)vma->vm_start, (void *)vma->vm_end, vma->vm_mm, (unsigned long )pgprot_val (vma->vm_page_prot), vma->anon_vma, vma->vm_ops,vma_start_pgoff (vma), vma->vm_file, vma->vm_private_data, -#ifdef CONFIG_PER_VMA_LOCKrefcount_read (&vma->vm_refcnt), -#endif vma->vm_flags, &vma->vm_flags); }EXPORT_SYMBOL (dump_vma);
mm/init-mm.c +0 -2
@@ -39,10 +39,8 @@ struct mm_struct init_mm = { .page_table_lock = __SPIN_LOCK_UNLOCKED(init_mm.page_table_lock), .arg_lock = __SPIN_LOCK_UNLOCKED(init_mm.arg_lock), .mmlist = LIST_HEAD_INIT(init_mm.mmlist), -#ifdef CONFIG_PER_VMA_LOCK .vma_writer_wait = __RCUWAIT_INITIALIZER(init_mm.vma_writer_wait), .mm_lock_seq = SEQCNT_ZERO(init_mm.mm_lock_seq), -#endif #ifdef CONFIG_SCHED_MM_CID .mm_cid.lock = __RAW_SPIN_LOCK_UNLOCKED(init_mm.mm_cid.lock),#endif
mm/memory.c +0 -2
@@ -6817,7 +6817,6 @@ static vm_fault_t sanitize_fault_flags(struct vm_area_struct *vma, !vma_is_cow_mapping(vma)))return VM_FAULT_SIGSEGV ; } -#ifdef CONFIG_PER_VMA_LOCK /* * Per-VMA locks can't be used with FAULT_FLAG_RETRY_NOWAIT because of * the assumption that lock is dropped on VM_FAULT_RETRY.
@@ -6826,7 +6825,6 @@ static vm_fault_t sanitize_fault_flags(struct vm_area_struct *vma, (FAULT_FLAG_VMA_LOCK | FAULT_FLAG_RETRY_NOWAIT)) == (FAULT_FLAG_VMA_LOCK | FAULT_FLAG_RETRY_NOWAIT)))return VM_FAULT_SIGSEGV ; -#endif return 0 ; }
mm/mmap_lock.c +2 -24
@@ -43,9 +43,6 @@ void __mmap_lock_do_trace_released(struct mm_struct *mm, bool write)EXPORT_SYMBOL (__mmap_lock_do_trace_released);#endif /* CONFIG_TRACING */ -#ifdef CONFIG_MMU -#ifdef CONFIG_PER_VMA_LOCK -/* State shared across __vma_[start, end]_exclude_readers. */ struct vma_exclude_readers_state {/* Input parameters. */
@@ -299,6 +296,8 @@ struct vm_area_struct *lock_vma_under_rcu(struct mm_struct *mm,MA_STATE (mas, &mm->mm_mt, address, address);struct vm_area_struct *vma; +if (!IS_ENABLED (CONFIG_MMU )) +return NULL ; retry:rcu_read_lock (); vma =mas_walk (&mas);
@@ -431,7 +430,6 @@ struct vm_area_struct *lock_next_vma(struct mm_struct *mm,return vma; } -#endif /* CONFIG_PER_VMA_LOCK */ #ifdef CONFIG_LOCK_MM_AND_FIND_VMA #include <linux/extable.h>
@@ -548,23 +546,3 @@ struct vm_area_struct *lock_mm_and_find_vma(struct mm_struct *mm, return NULL; }#endif /* CONFIG_LOCK_MM_AND_FIND_VMA */ - -#else /* CONFIG_MMU */ - -/* -* At least xtensa ends up having protection faults even with no -* MMU.. No stack expansion, at least. -*/ -struct vm_area_struct *lock_mm_and_find_vma (struct mm_struct *mm, -unsigned long addr,struct pt_regs *regs) -{ -struct vm_area_struct *vma; - -mmap_read_lock (mm); - vma =vma_lookup (mm, addr); -if (!vma) -mmap_read_unlock (mm); -return vma; -} - -#endif /* CONFIG_MMU */
mm/pagewalk.c +0 -2
@@ -444,7 +444,6 @@ static inline void process_mm_walk_lock(struct mm_struct *mm,static inline void process_vma_walk_lock (struct vm_area_struct *vma,enum page_walk_lock walk_lock) { -#ifdef CONFIG_PER_VMA_LOCK switch (walk_lock) {case PGWALK_WRLOCK :vma_start_write (vma);
@@ -459,7 +458,6 @@ static inline void process_vma_walk_lock(struct vm_area_struct *vma,/* PGWALK_RDLOCK is handled by process_mm_walk_lock */ break ; } -#endif }/*
mm/rmap.c +0 -2
@@ -260,11 +260,9 @@ static void check_anon_vma_clone(struct vm_area_struct *dst,/* For the anon_vma to be compatible, it can only be singular. */ VM_WARN_ON_ONCE (operation ==VMA_OP_MERGE_UNFAULTED && !list_is_singular (&src->anon_vma_chain)); -#ifdef CONFIG_PER_VMA_LOCK /* Only merging an unfaulted VMA leaves the destination attached. */ VM_WARN_ON_ONCE (operation !=VMA_OP_MERGE_UNFAULTED &&vma_is_attached (dst)); -#endif } static void maybe_reuse_anon_vma(struct vm_area_struct *dst,
mm/userfaultfd.c +0 -55
79 lines not on this page · raw message
rust/kernel/mm.rs +11 -21
@@ -170,30 +170,20 @@ pub unsafe fn from_raw<'a>(ptr: *const bindings::mm_struct) -> &'a MmWithUser {/// /// This is an optimistic trylock operation, so it may fail if there is contention. In that /// case, you should fall back to taking the mmap read lock. -/// -/// When per-vma locks are disabled, this always returns `None`. #[inline] pub fn lock_vma_under_rcu (&self, vma_addr:usize ) ->Option <VmaReadGuard <'_>> { -#[cfg( CONFIG_PER_VMA_LOCK )] - { -// SAFETY: Calling `bindings::lock_vma_under_rcu` is always okay given an mm where -// `mm_users` is non-zero. -let vma =unsafe { bindings::lock_vma_under_rcu (self.as_raw (), vma_addr) }; -if !vma.is_null () { -return Some (VmaReadGuard { -// SAFETY: If `lock_vma_under_rcu` returns a non-null ptr, then it points at a -// valid vma. The vma is stable for as long as the vma read lock is held. - vma:unsafe {VmaRef ::from_raw (vma) }, - _nts:NotThreadSafe , - }); - } +// SAFETY: Calling `bindings::lock_vma_under_rcu` is always okay given an mm where +// `mm_users` is non-zero. +let vma =unsafe { bindings::lock_vma_under_rcu (self.as_raw (), vma_addr) }; +if vma.is_null () { +return None ; } - -// Silence warnings about unused variables. -#[cfg(not( CONFIG_PER_VMA_LOCK ))] -let _ = vma_addr; - -None +Some (VmaReadGuard { +// SAFETY: If `lock_vma_under_rcu` returns a non-null ptr, then it points at a +// valid vma. The vma is stable for as long as the vma read lock is held. + vma:unsafe {VmaRef ::from_raw (vma) }, + _nts:NotThreadSafe , + }) }/// Lock the mmap read lock.
tools/testing/vma/include/dup.h +1 -4
@@ -560,7 +560,6 @@ struct vm_area_struct {vma_flags_t flags; }; -#ifdef CONFIG_PER_VMA_LOCK /* * Can only be written (using WRITE_ONCE()) while holding both: * - mmap_lock (in write mode)
@@ -576,7 +575,7 @@ struct vm_area_struct {* slowpath. */ unsigned int vm_lock_seq; -#endif +unsigned int __vm_anon_pgoff_lo;/*
@@ -610,10 +609,8 @@ struct vm_area_struct {#ifdef CONFIG_NUMA_BALANCING struct vma_numab_state *numab_state;/* NUMA Balancing state */ #endif -#ifdef CONFIG_PER_VMA_LOCK /* Unstable RCU readers are allowed to read this. */ refcount_t vm_refcnt; -#endif #ifdef CONFIG_64BIT unsigned int __vm_anon_pgoff_hi;#endif
tools/testing/vma/vma_internal.h +0 -1
@@ -15,7 +15,6 @@#include <stdlib.h> #define CONFIG_MMU 1 -#define CONFIG_PER_VMA_LOCK 1#ifdef __CONCAT#undef __CONCAT
Signature · 1 line
2.55.0.966.g6673acef38-goog
Re: [PATCH v7 1/5] mm: Make per-VMA locks available universally
Quoted 24 lines · Suren Baghdasaryan
From: Dave Hansen <[email protected]>
The per-VMA locks have been around for several years. They've had some bugs worked out of them and have seen quite wide use. However, they are still only available when architectures explicitly enable them. Remove the conditional compilation around the per-VMA locks, making them available on all architectures and configs.
The approach up to now seemed to be to add
ARCH_SUPPORTS_PER_VMA_LOCKwhen the architecture started using per-VMA locks in the fault handler. But, contrary to the naming, theKconfigoption does not really indicate whether the architecture supports per-VMA locks or not. It is more of a marker for whether the architecture is likely to benefit from per-VMA locks.To me, the most important thing side-effect of universal availability is letting per-VMA locks be used in SMP=n configs. This lets us use per-VMA locking in all x86 code without fallbacks.
Overall, this just generally makes the kernel simpler. Just look at the diffstat. It also opens the door to users that want to use the per-VMA locks in common code. Doing *that* brings additional simplifications.
Quoted 6 lines
The downside of this is adding some fields to
vm_area_structandmm_struct. There are likely ways to optimize this, especially for things like SMP=n configs. For now, do the simplest thing: use the same implementation everywhere.== Considerations for NOMMU config ==
Echoing what I commented on v6 after v7 was already sent (not realizing there was a v7 after my inbox got flooded)
Maybe the NOMMU "support" could have been had in a separate prep patch (where a lot of the description below could have been moved), making the patch itself just mostly a removal of code.
Signature · 3 lines
Cheers, David
Suren Baghdasaryan
maintainerRe: [PATCH v7 1/5] mm: Make per-VMA locks available universally
Quoted 31 lines · David Hildenbrand (Arm)
Quoted 29 lines · Suren Baghdasaryan
From: Dave Hansen <[email protected]>
The per-VMA locks have been around for several years. They've had some bugs worked out of them and have seen quite wide use. However, they are still only available when architectures explicitly enable them. Remove the conditional compilation around the per-VMA locks, making them available on all architectures and configs.
The approach up to now seemed to be to add
ARCH_SUPPORTS_PER_VMA_LOCKwhen the architecture started using per-VMA locks in the fault handler. But, contrary to the naming, theKconfigoption does not really indicate whether the architecture supports per-VMA locks or not. It is more of a marker for whether the architecture is likely to benefit from per-VMA locks.To me, the most important thing side-effect of universal availability is letting per-VMA locks be used in SMP=n configs. This lets us use per-VMA locking in all x86 code without fallbacks.
Overall, this just generally makes the kernel simpler. Just look at the diffstat. It also opens the door to users that want to use the per-VMA locks in common code. Doing *that* brings additional simplifications.
The downside of this is adding some fields to
vm_area_structandmm_struct. There are likely ways to optimize this, especially for things like SMP=n configs. For now, do the simplest thing: use the same implementation everywhere.
Quoted 8 lines
Quoted 1 line
== Considerations for NOMMU config ==
Echoing what I commented on v6 after v7 was already sent (not realizing there was a v7 after my inbox got flooded)
Maybe the NOMMU "support" could have been had in a separate prep patch (where a lot of the description below could have been moved), making the patch itself just mostly a removal of code.
Yeah, that would make reviewing indeed easier. I guess current structure is the result of me adding NOMMU pieces to the pre-existing patchset. I don't think it's worth refactoring but I can do that if you insist.
Quoted 3 lines
Thanks!
Quoted 5 lines
--
Cheers,David
David Hildenbrand (Arm)
maintainerRe: [PATCH v7 1/5] mm: Make per-VMA locks available universally
Quoted 38 lines · Suren Baghdasaryan
Quoted 36 lines · David Hildenbrand (Arm)
Quoted 24 lines · Suren Baghdasaryan
From: Dave Hansen <[email protected]>
The per-VMA locks have been around for several years. They've had some bugs worked out of them and have seen quite wide use. However, they are still only available when architectures explicitly enable them. Remove the conditional compilation around the per-VMA locks, making them available on all architectures and configs.
The approach up to now seemed to be to add
ARCH_SUPPORTS_PER_VMA_LOCKwhen the architecture started using per-VMA locks in the fault handler. But, contrary to the naming, theKconfigoption does not really indicate whether the architecture supports per-VMA locks or not. It is more of a marker for whether the architecture is likely to benefit from per-VMA locks.To me, the most important thing side-effect of universal availability is letting per-VMA locks be used in SMP=n configs. This lets us use per-VMA locking in all x86 code without fallbacks.
Overall, this just generally makes the kernel simpler. Just look at the diffstat. It also opens the door to users that want to use the per-VMA locks in common code. Doing *that* brings additional simplifications.
Quoted 6 lines
The downside of this is adding some fields to
vm_area_structandmm_struct. There are likely ways to optimize this, especially for things like SMP=n configs. For now, do the simplest thing: use the same implementation everywhere.== Considerations for NOMMU config ==
Echoing what I commented on v6 after v7 was already sent (not realizing there was a v7 after my inbox got flooded)
Quoted 8 lines
Quoted 3 lines
Maybe the NOMMU "support" could have been had in a separate prep patch (where a lot of the description below could have been moved), making the patch itself just mostly a removal of code.
Yeah, that would make reviewing indeed easier. I guess current structure is the result of me adding NOMMU pieces to the pre-existing patchset. I don't think it's worth refactoring but I can do that if you insist.
I won't insist, it was just commenting because it would cleanly separate the NOMMU oddity from just unlocking it in common code and removing the leftovers :)
Signature · 3 lines
Cheers, David
Suren Baghdasaryan
maintainerPATCH v7 2/5 binder: Make shrinker rely solely on per-VMA lock
In reply to Suren BaghdasaryanFrom: Dave Hansen <[email protected]>
tl;dr: lock_vma_under_rcu() is already a trylock. No need to do both it and mmap_read_trylock().
Long Version:
== Background ==
Historically, binder used an mmap_read_trylock() in its shrinker code. This ensures that reclaim is not blocked on an mmap_lock. Commit 95bc2d4a9020 ("binder: use per-vma lock in page reclaiming") added support for the per-VMA lock, but left mmap_read_trylock() as a fallback.
This was presumably because the per-VMA locking can fail for several reasons and most (all?) lock_vma_under_rcu() callers have a fallback to mmap_read_trylock().
== Problem ==
The fallback is not worth the complexity here. lock_vma_under_rcu() is essentially already a non-blocking trylock. The main reason it fails is also the reason mmap_read_trylock() fails: something is holding mmap_write_lock().
The only remedy for a collision with mmap_write_lock() is to wait, which this code can not do. So the "fallback" after lock_vma_under_rcu() failure is not really a fallback: it is really likely to just be retrying in vain. That retry in an of itself isn't horrible. But it adds complexity.
== Solution ==
Now that per-VMA locks are universally available, lock_vma_under_rcu() will not persistently fail. Rely on it alone and simplify the code. The removal of the fallback does not affect NOMMU case because binder driver depends on CONFIG_MMU.
While at it we also make the handling of the cases where the original binder VMA is gone consistent. There are two cases to consider when Binder VMA is gone:
1. there is no VMA at that location anymore.
2. there is now another unrelated VMA at that location.
Before this change we handle case 1 by having the shrinker proceed to free the page, and just skip the zap_vma_range() call. And we handle case 2 by having the shrinker return LRU_SKIP. While either behavior is acceptable, we need to handle them in a consistent way. Handle both cases by freeing the page without touching the VMA (skipping the zap_vma_range()).
Full disclosure: I originally tried to do this with lock_vma_under_rcu_wait(), but it did not fit well with the mmap_lock trylock semantics. Claude caught this in a review and suggested the approach in this path. It seemed sane to me. So, Suggesed-by: Claude, I guess.
--- drivers/android/binder_alloc.c | 46 ++++++++++++++++------------------ 1 file changed, 21 insertions(+), 25 deletions(-)
drivers/android/binder_alloc.c +21 -25
@@ -1142,7 +1142,6 @@ enum lru_status binder_alloc_free_page(struct list_head *item,struct vm_area_struct *vma;struct page *page_to_free;unsigned long page_addr; -int mm_locked =0 ;size_t index;if (!mmget_not_zero (mm))
@@ -1151,27 +1150,25 @@ enum lru_status binder_alloc_free_page(struct list_head *item, index = mdata->page_index; page_addr = alloc->vm_start + index *PAGE_SIZE ; -/* attempt per-vma lock first */ +/* +* Attempt per-vma lock. This is essentially a +* "trylock". It can fail even if the VMA exists +* for 'page_addr'. +*/ vma =lock_vma_under_rcu (mm, page_addr);if (!vma) { -/* fall back to mmap_lock */ -if (!mmap_read_trylock (mm)) - goto err_mmap_read_lock_failed; - mm_locked =1 ; - vma =vma_lookup (mm, page_addr); +/* +* If the vma exists, we can't continue because we cannot +* remove the page from the vma. However, if the vma was +* unmapped, it's okay to continue. +*/ +if (binder_alloc_is_mapped (alloc)) + goto err_vma_lock_failed; }if (!mutex_trylock (&alloc->mutex)) goto err_get_alloc_mutex_failed; -/* -* Since a binder_alloc can only be mapped once, we ensure -* the vma corresponds to this mapping by checking whether -* the binder_alloc is still mapped. -*/ -if (vma && !binder_alloc_is_mapped (alloc)) - goto err_invalid_vma; -trace_binder_unmap_kernel_start (alloc, index); page_to_free = alloc->pages[index];
@@ -1182,7 +1179,12 @@ enum lru_status binder_alloc_free_page(struct list_head *item,list_lru_isolate (lru, item);spin_unlock (&lru->lock); -if (vma) { +/* +* Since a binder_alloc can only be mapped once, we ensure +* the vma corresponds to this mapping by checking whether +* the binder_alloc is still mapped. +*/ +if (vma &&binder_alloc_is_mapped (alloc)) {trace_binder_unmap_user_start (alloc, index);zap_vma_range (vma, page_addr,PAGE_SIZE );
@@ -1191,23 +1193,17 @@ enum lru_status binder_alloc_free_page(struct list_head *item, }mutex_unlock (&alloc->mutex); -if (mm_locked) -mmap_read_unlock (mm); -else +if (vma)vma_end_read (vma);mmput_async (mm);binder_free_page (page_to_free);return LRU_REMOVED_RETRY ; -err_invalid_vma: -mutex_unlock (&alloc->mutex); err_get_alloc_mutex_failed: -if (mm_locked) -mmap_read_unlock (mm); -else +if (vma)vma_end_read (vma); -err_mmap_read_lock_failed: +err_vma_lock_failed:mmput_async (mm); err_mmget:return LRU_SKIP ;
Signature · 1 line
2.55.0.966.g6673acef38-goog
Greg KH
maintainerRe: [PATCH v7 2/5] binder: Make shrinker rely solely on per-VMA lock
Quoted 77 lines · Suren Baghdasaryan
From: Dave Hansen <[email protected]>
tl;dr:
lock_vma_under_rcu()is already a trylock. No need to do both it andmmap_read_trylock().Long Version:
== Background ==
Historically, binder used an
mmap_read_trylock()in its shrinker code. This ensures that reclaim is not blocked on anmmap_lock. Commit95bc2d4a9020("binder: use per-vma lock in page reclaiming") added support for the per-VMA lock, but leftmmap_read_trylock()as a fallback.This was presumably because the per-VMA locking can fail for several reasons and most (all?)
lock_vma_under_rcu()callers have a fallback tommap_read_trylock().== Problem ==
The fallback is not worth the complexity here.
lock_vma_under_rcu()is essentially already a non-blocking trylock. The main reason it fails is also the reasonmmap_read_trylock()fails: something is holdingmmap_write_lock().The only remedy for a collision with
mmap_write_lock()is to wait, which this code can not do. So the "fallback" afterlock_vma_under_rcu()failure is not really a fallback: it is really likely to just be retrying in vain. That retry in an of itself isn't horrible. But it adds complexity.== Solution ==
Now that per-VMA locks are universally available,
lock_vma_under_rcu()will not persistently fail. Rely on it alone and simplify the code. The removal of the fallback does not affect NOMMU case because binder driver depends onCONFIG_MMU.While at it we also make the handling of the cases where the original binder VMA is gone consistent. There are two cases to consider when Binder VMA is gone:
1. there is no VMA at that location anymore.
2. there is now another unrelated VMA at that location.Before this change we handle case 1 by having the shrinker proceed to free the page, and just skip the
zap_vma_range()call. And we handle case 2 by having the shrinker returnLRU_SKIP. While either behavior is acceptable, we need to handle them in a consistent way. Handle both cases by freeing the page without touching the VMA (skipping thezap_vma_range()).Full disclosure: I originally tried to do this with
lock_vma_under_rcu_wait(), but it did not fit well with themmap_locktrylock semantics. Claude caught this in a review and suggested the approach in this path. It seemed sane to me. So, Suggesed-by: Claude, I guess.Signed-off-by Dave Hansen <[email protected]>Cc Andrew Morton <[email protected]>Cc Liam R. Howlett <[email protected]>Cc Vlastimil Babka <[email protected]>Cc Shakeel Butt <[email protected]>Cc Greg Kroah-Hartman <[email protected]>Cc Arve Hjønnevåg <[email protected]>Cc Todd Kjos <[email protected]>Cc Christian Brauner <[email protected]>Cc Carlos Llamas <[email protected]>Cc Alice Ryhl <[email protected]>Cc David S. Miller <[email protected]>Cc David Ahern <[email protected]>Signed-off-by Suren Baghdasaryan <[email protected]>Reviewed-by Alice Ryhl <[email protected]>Acked-by Carlos Llamas <[email protected]>
Quoted 2 lines
--- drivers/android/binder_alloc.c | 46 ++++++++++++++++------------------
This file will be going away (along with the other binder .c files) for 7.4-rc1, so I don't object to this getting into other trees now, just note that it will not be needed if you wait a release cycle :)
thanks,
greg k-h
Suren Baghdasaryan
maintainerPATCH v7 3/5 mm: Add RCU-based VMA lookup helper that waits for writers
In reply to Suren BaghdasaryanFrom: Dave Hansen <[email protected]>
There are basically two parallel ways to look up a VMA: the traditional way, which is protected by mmap_read_lock, and the RCU-based per-VMA lock way which is based on RCU and refcounts. However, per-VMA locks will fail if the lock is help by a writer and therefore never waits. In a number of places we need to wait for the lock and it's done by falling back to mmap_read_lock, locking the VMA and releasing the mmap_lock once VMA is locked.
Add vma_start_read_unlocked() - a variant of the RCU-based lookup that waits for writers. This is basically the same as the existing RCU-based lookup, but on a failure to lock it temporarily takes mmap_lock for read and waits for writers to finish before locking the VMA, dropping the mmap_read_lock and returning the locked VMA. This has some advantages:
1. Callers do not need to have a fallback path for when they collide with writers.
2. Its fast path does not require taking mmap_lock for read.
Basically, when applied correctly, this approach results in faster *and* simpler code.
While at it, fix the comments for vma_start_read_locked(), vma_start_read_locked_nested(), and uffd_lock_vma().
--- include/linux/mmap_lock.h | 19 +++++++++++++++---- mm/mmap_lock.c | 35 +++++++++++++++++++++++++++++++++++ mm/userfaultfd.c | 6 ++++-- 3 files changed, 54 insertions(+), 6 deletions(-)
include/linux/mmap_lock.h +15 -4
@@ -228,10 +228,14 @@ static inline void vma_refcount_put(struct vm_area_struct *vma) }/* -* Use only while holding mmap read lock which guarantees that locking will not -* fail (nobody can concurrently write-lock the vma). vma_start_read() should +* Use only while holding mmap read lock which guarantees that vma lock is not +* contended (nobody can concurrently write-lock the vma). vma_start_read() should * not be used in such cases because it might fail due to mm_lock_seq overflow. * This functionality is used to obtain vma read lock and drop the mmap read lock. +* +* VMA can't be detached while we are holding mmap lock, therefore in practice this +* function can fail only when there are so many readers that vm_refcnt overflows. +* The failure case is very unlikely and is already annotated as such internally. */ static inline bool vma_start_read_locked_nested (struct vm_area_struct *vma,int subclass) {
@@ -247,16 +251,23 @@ static inline bool vma_start_read_locked_nested(struct vm_area_struct *vma, int }/* -* Use only while holding mmap read lock which guarantees that locking will not -* fail (nobody can concurrently write-lock the vma). vma_start_read() should +* Use only while holding mmap read lock which guarantees that vma lock is not +* contended (nobody can concurrently write-lock the vma). vma_start_read() should * not be used in such cases because it might fail due to mm_lock_seq overflow. * This functionality is used to obtain vma read lock and drop the mmap read lock. +* +* VMA can't be detached while we are holding mmap lock, therefore in practice this +* function can fail only when there are so many readers that vm_refcnt overflows. +* The failure case is very unlikely and is already annotated as such internally. */ static inline bool vma_start_read_locked (struct vm_area_struct *vma) {return vma_start_read_locked_nested (vma,0 ); } +struct vm_area_struct *vma_start_read_unlocked (struct mm_struct *mm, +unsigned long address); +static inline void vma_end_read (struct vm_area_struct *vma) {vma_refcount_put (vma);
mm/mmap_lock.c +35 -0
@@ -340,6 +340,41 @@ struct vm_area_struct *lock_vma_under_rcu(struct mm_struct *mm,return NULL ; } +/** +* vma_start_read_unlocked() - Find the VMA covering 'address' and read-lock it. +* @mm: the mm_struct of the address space to search +* @address: address that the vma should contain +* +* The fast path does not take mmap_lock. Waits for writers to finish if the +* VMA is being modified by taking mmap_lock. +* Use when mmap_lock is not held, otherwise use vma_start_read_locked(). +* Nothing prevents VMAs being unmapped/mapped before or after the VMA is +* looked up, if a stronger guarantee is required, take an mmap_lock. +* +* Return: If a VMA exists which spans @address, return that VMA, read-locked. +* If no VMA is mapped there or, very unlikely, a reference count overflow +* occurred, return NULL. +*/ +struct vm_area_struct *vma_start_read_unlocked (struct mm_struct *mm, +unsigned long address) +{ +struct vm_area_struct *vma; + +/* Fast path: return stable VMA covering 'address': */ + vma =lock_vma_under_rcu (mm, address); +if (vma) +return vma; + +/* Slow path: preclude VMA writers by temporarily getting mmap read lock. */ +mmap_read_lock (mm); + vma =vma_lookup (mm, address); +if (vma && !vma_start_read_locked (vma)) + vma =NULL ; +mmap_read_unlock (mm); + +return vma; +} +static struct vm_area_struct *lock_next_vma_under_mmap_lock (struct mm_struct *mm,struct vma_iterator *vmi,unsigned long from_addr)
mm/userfaultfd.c +4 -2
@@ -129,8 +129,10 @@ struct vm_area_struct *find_vma_and_prepare_anon(struct mm_struct *mm,* * Should be called without holding mmap_lock. * -* Return: A locked vma containing @address, -ENOENT if no vma is found, or -* -ENOMEM if anon_vma couldn't be allocated. +* Return: A locked vma containing @address, -ENOENT if no vma is found, +* -ENOMEM if anon_vma couldn't be allocated, or -EAGAIN if vma refcount +* overflow happened due to high number of readers and the caller should +* retry later. */ static struct vm_area_struct *uffd_lock_vma (struct mm_struct *mm,unsigned long address)
Signature · 1 line
2.55.0.966.g6673acef38-goog
Suren Baghdasaryan
maintainerPATCH v7 4/5 binder: Remove mmap_lock fallback
In reply to Suren BaghdasaryanFrom: Dave Hansen <[email protected]>
Previously, the per-VMA locking could fail in the face of writers which necessitate a fallback to mmap_lock. The new vma_start_read_unlocked() will wait for writers instead of failing.
Use the new helper. Wait for writers. Remove the fallback to mmap_lock.
--- drivers/android/binder/page_range.rs | 19 +++---------------- drivers/android/binder_alloc.c | 17 +++++------------ rust/kernel/mm.rs | 28 ++++++++++++++++++++++++++++ 3 files changed, 36 insertions(+), 28 deletions(-)
drivers/android/binder/page_range.rs +3 -16
@@ -439,22 +439,9 @@ unsafe fn use_page_slow(&self, i: usize) -> Result<()> {// workqueue. let mm =MmWithUser ::into_mmput_async (self.mm.mmget_not_zero ().ok_or (ESRCH )?); { -let vma_read; -let mmap_read; -let vma =if let Some (ret) = mm.lock_vma_under_rcu (vma_addr) { - vma_read = ret; -check_vma (&vma_read, self) - }else { - mmap_read = mm.mmap_read_lock (); - mmap_read - .vma_lookup (vma_addr) - .and_then (|vma|check_vma (vma, self)) - }; - -match vma { -Some (vma) => vma.vm_insert_page (user_page_addr, &new_page)?, -None =>return Err (ESRCH ), - } +let vma_read_guard = mm.vma_start_read_unlocked (vma_addr).ok_or (ESRCH )?; +let vma =check_vma (&vma_read_guard, self).ok_or (ESRCH )?; + vma.vm_insert_page (user_page_addr, &new_page)?; }let inner = self.lock.lock ();
drivers/android/binder_alloc.c +5 -12
@@ -259,21 +259,14 @@ static int binder_page_insert(struct binder_alloc *alloc,struct vm_area_struct *vma;int ret = -ESRCH ; -/* attempt per-vma lock first */ - vma =lock_vma_under_rcu (mm, addr); -if (vma) { -if (binder_alloc_is_mapped (alloc)) - ret =vm_insert_page (vma, addr, page); -vma_end_read (vma); + vma =vma_start_read_unlocked (mm, addr); +if (!vma)return ret; - } -/* fall back to mmap_lock */ -mmap_read_lock (mm); - vma =vma_lookup (mm, addr); -if (vma &&binder_alloc_is_mapped (alloc)) +if (binder_alloc_is_mapped (alloc)) ret =vm_insert_page (vma, addr, page); -mmap_read_unlock (mm); + +vma_end_read (vma);return ret; }
rust/kernel/mm.rs +28 -0
@@ -186,6 +186,34 @@ pub fn lock_vma_under_rcu(&self, vma_addr: usize) -> Option<VmaReadGuard<'_>> { }) } +/// Find the VMA covering 'address' and read-lock it. +/// +/// The fast path does not take mmap_lock. Waits for writers to finish if the +/// VMA is being modified by taking mmap_lock. +/// Use when mmap_lock is not held, otherwise use vma_start_read_locked(). +/// Nothing prevents VMAs being unmapped/mapped before or after the VMA is +/// looked up, if a stronger guarantee is required, take an mmap_lock. +/// +/// Return: If a VMA exists which spans @address, return that VMA, read-locked. +/// If no VMA is mapped there or, very unlikely, a reference count overflow +/// occurred, return NULL. +#[inline] +pub fn vma_start_read_unlocked (&self, vma_addr:usize ) ->Option <VmaReadGuard <'_>> { +// SAFETY: We may invoke `vma_start_read_unlocked` because we know this `mm` has non-zero +// `mm_users`. +let vma =unsafe { bindings::vma_start_read_unlocked (self.as_raw (), vma_addr) }; +if vma.is_null () { +return None ; + } +// INVARIANT: We just acquired the VMA read lock. +Some (VmaReadGuard { +// SAFETY: If `vma_start_read_unlocked` returns a non-null ptr, then it points at a +// valid vma. The vma is stable for as long as the vma read lock is held. + vma:unsafe {VmaRef ::from_raw (vma) }, + _nts:NotThreadSafe , + }) + } +/// Lock the mmap read lock. #[inline] pub fn mmap_read_lock (&self) ->MmapReadGuard <'_> {
Signature · 1 line
2.55.0.966.g6673acef38-goog
Suren Baghdasaryan
maintainerPATCH v7 5/5 tcp: Remove mmap_lock fallback path
In reply to Suren BaghdasaryanFrom: Dave Hansen <[email protected]>
Previously, the per-VMA locking could fail in the face of writers which necessitates a fallback to mmap_lock. The new vma_start_read_unlocked() will wait for writers instead of failing.
Use the new helper. Wait for writers. Remove the fallback to mmap_lock. The fallback removal does not affect NOMMU case because TCP_ZEROCOPY is gated on CONFIG_MMU.
This really is a nice cleanup. It removes the need to pass the lock state back and forth to find_tcp_vma().
--- net/ipv4/tcp.c | 31 +++++++++---------------------- 1 file changed, 9 insertions(+), 22 deletions(-)
net/ipv4/tcp.c +9 -22
@@ -2167,27 +2167,18 @@ static void tcp_zc_finalize_rx_tstamp(struct sock *sk, }static struct vm_area_struct *find_tcp_vma (struct mm_struct *mm, -unsigned long address, -bool *mmap_locked) +unsigned long address) { -struct vm_area_struct *vma =lock_vma_under_rcu (mm, address); +struct vm_area_struct *vma =vma_start_read_unlocked (mm, address); -if (vma) { -if (vma->vm_ops != &tcp_vm_ops) { -vma_end_read (vma); -return NULL ; - } - *mmap_locked = false; -return vma; - } +if (!vma) +return NULL ; -mmap_read_lock (mm); - vma =vma_lookup (mm, address); -if (!vma || vma->vm_ops != &tcp_vm_ops) { -mmap_read_unlock (mm); +if (vma->vm_ops != &tcp_vm_ops) { +vma_end_read (vma);return NULL ; } - *mmap_locked = true; +return vma; }
@@ -2208,7 +2199,6 @@ static int tcp_zerocopy_receive(struct sock *sk,u32 seq = tp->copied_seq;u32 total_bytes_to_map;int inq =tcp_inq (sk); -bool mmap_locked;int ret; zc->copybuf_len =0 ;
@@ -2233,7 +2223,7 @@ static int tcp_zerocopy_receive(struct sock *sk,return 0 ; } - vma =find_tcp_vma (current->mm, address, &mmap_locked); + vma =find_tcp_vma (current->mm, address);if (!vma)return -EINVAL ;
@@ -2315,10 +2305,7 @@ static int tcp_zerocopy_receive(struct sock *sk, zc, total_bytes_to_map); } out: -if (mmap_locked) -mmap_read_unlock (current->mm); -else -vma_end_read (vma); +vma_end_read (vma);/* Try to copy straggler data. */ if (!ret) copylen =tcp_zc_handle_leftover (zc, sk, skb, &seq, copybuf_len, tss);
Signature · 1 line
2.55.0.966.g6673acef38-goog
Paolo Abeni
maintainerRe: [PATCH v7 5/5] tcp: Remove mmap_lock fallback path
adding Eric.
Full series at:
https://lore.kernel.org/netdev/[email protected]/
Quoted 104 lines · Suren Baghdasaryan
From: Dave Hansen <[email protected]>
Previously, the per-VMA locking could fail in the face of writers which necessitates a fallback to
mmap_lock. The newvma_start_read_unlocked()will wait for writers instead of failing.Use the new helper. Wait for writers. Remove the fallback to
mmap_lock. The fallback removal does not affect NOMMU case becauseTCP_ZEROCOPYis gated onCONFIG_MMU.This really is a nice cleanup. It removes the need to pass the lock state back and forth to
find_tcp_vma().Signed-off-by Dave Hansen <[email protected]>Acked-by Lorenzo Stoakes <[email protected]>Tested-by [email protected]Cc Andrew Morton <[email protected]>Cc Liam R. Howlett <[email protected]>Cc Vlastimil Babka <[email protected]>Cc Shakeel Butt <[email protected]>Cc Greg Kroah-Hartman <[email protected]>Cc Arve Hjønnevåg <[email protected]>Cc Todd Kjos <[email protected]>Cc Christian Brauner <[email protected]>Cc Carlos Llamas <[email protected]>Cc Alice Ryhl <[email protected]>Cc David S. Miller <[email protected]>Cc David Ahern <[email protected]>Signed-off-by Suren Baghdasaryan <[email protected]>--- net/ipv4/tcp.c | 31 +++++++++---------------------- 1 file changed, 9 insertions(+), 22 deletions(-)net/ipv4/tcp.c +9
-22@@ -2167,27 +2167,18 @@ static void tcp_zc_finalize_rx_tstamp(struct sock *sk, }static struct vm_area_struct *find_tcp_vma (struct mm_struct *mm, -unsigned long address, -bool *mmap_locked) +unsigned long address) { -struct vm_area_struct *vma =lock_vma_under_rcu (mm, address); +struct vm_area_struct *vma =vma_start_read_unlocked (mm, address); -if (vma) { -if (vma->vm_ops != &tcp_vm_ops) { -vma_end_read (vma); -return NULL ; - } - *mmap_locked = false; -return vma; - } +if (!vma) +return NULL ; -mmap_read_lock (mm); - vma =vma_lookup (mm, address); -if (!vma || vma->vm_ops != &tcp_vm_ops) { -mmap_read_unlock (mm); +if (vma->vm_ops != &tcp_vm_ops) { +vma_end_read (vma);return NULL ; } - *mmap_locked = true; +return vma; }@@ -2208,7 +2199,6 @@ static int tcp_zerocopy_receive(struct sock *sk,u32 seq = tp->copied_seq;u32 total_bytes_to_map;int inq =tcp_inq (sk); -bool mmap_locked;int ret; zc->copybuf_len =0 ;@@ -2233,7 +2223,7 @@ static int tcp_zerocopy_receive(struct sock *sk,return 0 ; } - vma =find_tcp_vma (current->mm, address, &mmap_locked); + vma =find_tcp_vma (current->mm, address);if (!vma)return -EINVAL ;@@ -2315,10 +2305,7 @@ static int tcp_zerocopy_receive(struct sock *sk, zc, total_bytes_to_map); } out: -if (mmap_locked) -mmap_read_unlock (current->mm); -else -vma_end_read (vma); +vma_end_read (vma);/* Try to copy straggler data. */ if (!ret) copylen =tcp_zc_handle_leftover (zc, sk, skb, &seq, copybuf_len, tss);