[Git][ghc/ghc][wip/T23400] rts: Ensure that pinned allocations respect block size
Ben Gamari (@bgamari)
gitlab at gitlab.haskell.org
Wed May 24 19:59:22 UTC 2023
Ben Gamari pushed to branch wip/T23400 at Glasgow Haskell Compiler / GHC
Commits:
26480033 by Ben Gamari at 2023-05-24T15:58:32-04:00
rts: Ensure that pinned allocations respect block size
Previously, it was possible for pinned, aligned allocation requests to
allocate beyond the end of the pinned accumulator block. Specifically,
we failed to account for the padding needed to achieve the requested
alignment in the "large object" check. With large alignment requests,
this can result in the allocator using the capability's pinned object
accumulator block to service a request which is larger than
`PINNED_EMPTY_SIZE`.
To fix this we reorganize `allocatePinned` to consistently account for
the alignment padding in all large object checks. This is a bit subtle
as we must handle the case of a small allocation request filling the
accumulator block, as well as large requests.
Fixes #23400.
- - - - -
1 changed file:
- rts/sm/Storage.c
Changes:
=====================================
rts/sm/Storage.c
=====================================
@@ -1231,6 +1231,74 @@ allocateMightFail (Capability *cap, W_ n)
*/
#define MEMSET_SLOP_W(p, val, len_w) memset(p, val, (len_w) * sizeof(W_))
+/**
+ * Finish the capability's current pinned object accumulator block
+ * (cap->pinned_object_block), if any, and start a new one.
+ */
+static bdescr *
+start_new_pinned_block(Capability *cap)
+{
+ bdescr *bd = cap->pinned_object_block;
+
+ // stash the old block on cap->pinned_object_blocks. On the
+ // next GC cycle these objects will be moved to
+ // g0->large_objects.
+ if (bd != NULL) {
+ // add it to the allocation stats when the block is full
+ finishedNurseryBlock(cap, bd);
+ dbl_link_onto(bd, &cap->pinned_object_blocks);
+ }
+
+ // We need to find another block. We could just allocate one,
+ // but that means taking a global lock and we really want to
+ // avoid that (benchmarks that allocate a lot of pinned
+ // objects scale really badly if we do this).
+ //
+ // See Note [Sources of Block Level Fragmentation]
+ // for a more complete history of this section.
+ bd = cap->pinned_object_empty;
+ if (bd == NULL) {
+ // The pinned block list is empty: allocate a fresh block (we can't fail
+ // here).
+ ACQUIRE_SM_LOCK;
+ bd = allocNursery(cap->node, NULL, PINNED_EMPTY_SIZE);
+ RELEASE_SM_LOCK;
+ }
+
+ // Bump up the nursery pointer to avoid the pathological situation
+ // where a program is *only* allocating pinned objects.
+ // T4018 fails without this safety.
+ // This has the effect of counting a full pinned block in the same way
+ // as a full nursery block, so GCs will be triggered at the same interval
+ // if you are only allocating pinned data compared to normal allocations
+ // via allocate().
+ bdescr *nbd = cap->r.rCurrentNursery->link;
+ if (nbd != NULL){
+ newNurseryBlock(nbd);
+ cap->r.rCurrentNursery->link = nbd->link;
+ if (nbd->link != NULL) {
+ nbd->link->u.back = cap->r.rCurrentNursery;
+ }
+ dbl_link_onto(nbd, &cap->r.rNursery->blocks);
+ // Important for accounting purposes
+ if (cap->r.rCurrentAlloc){
+ finishedNurseryBlock(cap, cap->r.rCurrentAlloc);
+ }
+ cap->r.rCurrentAlloc = nbd;
+ }
+
+ cap->pinned_object_empty = bd->link;
+ newNurseryBlock(bd);
+ if (bd->link != NULL) {
+ bd->link->u.back = cap->pinned_object_empty;
+ }
+ initBdescr(bd, g0, g0);
+
+ cap->pinned_object_block = bd;
+ bd->flags = BF_PINNED | BF_LARGE | BF_EVACUATED;
+ return bd;
+}
+
/* ---------------------------------------------------------------------------
Allocate a fixed/pinned object.
@@ -1258,135 +1326,76 @@ allocateMightFail (Capability *cap, W_ n)
StgPtr
allocatePinned (Capability *cap, W_ n /*words*/, W_ alignment /*bytes*/, W_ align_off /*bytes*/)
{
- StgPtr p;
- bdescr *bd;
-
// Alignment and offset have to be a power of two
- ASSERT(alignment && !(alignment & (alignment - 1)));
- ASSERT(alignment >= sizeof(W_));
-
- ASSERT(!(align_off & (align_off - 1)));
+ CHECK(alignment && !(alignment & (alignment - 1)));
+ CHECK(!(align_off & (align_off - 1)));
+ // We don't support sub-word alignments
+ CHECK(alignment >= sizeof(W_));
+
+ bdescr *bd = cap->pinned_object_block;
+ if (bd == NULL) {
+ bd = start_new_pinned_block(cap);
+ }
const StgWord alignment_w = alignment / sizeof(W_);
+ W_ off_w = ALIGN_WITH_OFF_W(bd->free, alignment, align_off);
+
+ // If the request is is smaller than LARGE_OBJECT_THRESHOLD then
+ // allocate into the pinned object accumulator.
+ if (n + off_w < LARGE_OBJECT_THRESHOLD/sizeof(W_)) {
+ // If the current pinned object block isn't large enough to hold the new
+ // object, get a new one.
+ if ((bd->free + off_w + n) > (bd->start + BLOCK_SIZE_W)) {
+ bd = start_new_pinned_block(cap);
+
+ // The pinned_object_block remains attached to the capability
+ // until it is full, even if a GC occurs. We want this
+ // behaviour because otherwise the unallocated portion of the
+ // block would be forever slop, and under certain workloads
+ // (allocating a few ByteStrings per GC) we accumulate a lot
+ // of slop.
+ //
+ // So, the pinned_object_block is initially marked
+ // BF_EVACUATED so the GC won't touch it. When it is full,
+ // we place it on the large_objects list, and at the start of
+ // the next GC the BF_EVACUATED flag will be cleared, and the
+ // block will be promoted as usual (if anything in it is
+ // live).
+
+ off_w = ALIGN_WITH_OFF_W(bd->free, alignment, align_off);
+ }
- // If the request is for a large object, then allocate()
- // will give us a pinned object anyway.
- if (n >= LARGE_OBJECT_THRESHOLD/sizeof(W_)) {
- // For large objects we don't bother optimizing the number of words
- // allocated for alignment reasons. Here we just allocate the maximum
- // number of extra words we could possibly need to satisfy the alignment
- // constraint.
- p = allocateMightFail(cap, n + alignment_w - 1);
- if (p == NULL) {
- return NULL;
- } else {
- Bdescr(p)->flags |= BF_PINNED;
- W_ off_w = ALIGN_WITH_OFF_W(p, alignment, align_off);
+ // N.B. it is important that we account for the alignment padding
+ // when determining large-object-ness, lest we may over-fill the
+ // block. See #23400.
+ if (n + off_w < LARGE_OBJECT_THRESHOLD/sizeof(W_)) {
+ StgPtr p = bd->free;
MEMSET_SLOP_W(p, 0, off_w);
+ n += off_w;
p += off_w;
- MEMSET_SLOP_W(p + n, 0, alignment_w - off_w - 1);
+ bd->free += n;
+ ASSERT(bd->free <= bd->start + bd->blocks * BLOCK_SIZE_W);
+ accountAllocation(cap, n);
return p;
}
}
- bd = cap->pinned_object_block;
-
- W_ off_w = 0;
-
- if(bd)
- off_w = ALIGN_WITH_OFF_W(bd->free, alignment, align_off);
-
- // If we don't have a block of pinned objects yet, or the current
- // one isn't large enough to hold the new object, get a new one.
- if (bd == NULL || (bd->free + off_w + n) > (bd->start + BLOCK_SIZE_W)) {
-
- // stash the old block on cap->pinned_object_blocks. On the
- // next GC cycle these objects will be moved to
- // g0->large_objects.
- if (bd != NULL) {
- // add it to the allocation stats when the block is full
- finishedNurseryBlock(cap, bd);
- dbl_link_onto(bd, &cap->pinned_object_blocks);
- }
-
- // We need to find another block. We could just allocate one,
- // but that means taking a global lock and we really want to
- // avoid that (benchmarks that allocate a lot of pinned
- // objects scale really badly if we do this).
- //
- // See Note [Sources of Block Level Fragmentation]
- // for a more complete history of this section.
- bd = cap->pinned_object_empty;
- if (bd == NULL) {
- // The pinned block list is empty: allocate a fresh block (we can't fail
- // here).
- ACQUIRE_SM_LOCK;
- bd = allocNursery(cap->node, NULL, PINNED_EMPTY_SIZE);
- RELEASE_SM_LOCK;
- }
-
- // Bump up the nursery pointer to avoid the pathological situation
- // where a program is *only* allocating pinned objects.
- // T4018 fails without this safety.
- // This has the effect of counting a full pinned block in the same way
- // as a full nursery block, so GCs will be triggered at the same interval
- // if you are only allocating pinned data compared to normal allocations
- // via allocate().
- bdescr * nbd;
- nbd = cap->r.rCurrentNursery->link;
- if (nbd != NULL){
- newNurseryBlock(nbd);
- cap->r.rCurrentNursery->link = nbd->link;
- if (nbd->link != NULL) {
- nbd->link->u.back = cap->r.rCurrentNursery;
- }
- dbl_link_onto(nbd, &cap->r.rNursery->blocks);
- // Important for accounting purposes
- if (cap->r.rCurrentAlloc){
- finishedNurseryBlock(cap, cap->r.rCurrentAlloc);
- }
- cap->r.rCurrentAlloc = nbd;
- }
-
-
- cap->pinned_object_empty = bd->link;
- newNurseryBlock(bd);
- if (bd->link != NULL) {
- bd->link->u.back = cap->pinned_object_empty;
- }
- initBdescr(bd, g0, g0);
-
- cap->pinned_object_block = bd;
- bd->flags = BF_PINNED | BF_LARGE | BF_EVACUATED;
-
- // The pinned_object_block remains attached to the capability
- // until it is full, even if a GC occurs. We want this
- // behaviour because otherwise the unallocated portion of the
- // block would be forever slop, and under certain workloads
- // (allocating a few ByteStrings per GC) we accumulate a lot
- // of slop.
- //
- // So, the pinned_object_block is initially marked
- // BF_EVACUATED so the GC won't touch it. When it is full,
- // we place it on the large_objects list, and at the start of
- // the next GC the BF_EVACUATED flag will be cleared, and the
- // block will be promoted as usual (if anything in it is
- // live).
-
- off_w = ALIGN_WITH_OFF_W(bd->free, alignment, align_off);
+ // Otherwise handle the request as a large object
+ // For large objects we don't bother optimizing the number of words
+ // allocated for alignment reasons. Here we just allocate the maximum
+ // number of extra words we could possibly need to satisfy the alignment
+ // constraint.
+ StgPtr p = allocateMightFail(cap, n + alignment_w - 1);
+ if (p == NULL) {
+ return NULL;
+ } else {
+ Bdescr(p)->flags |= BF_PINNED;
+ off_w = ALIGN_WITH_OFF_W(p, alignment, align_off);
+ MEMSET_SLOP_W(p, 0, off_w);
+ p += off_w;
+ MEMSET_SLOP_W(p + n, 0, alignment_w - off_w - 1);
+ return p;
}
-
- p = bd->free;
-
- MEMSET_SLOP_W(p, 0, off_w);
-
- n += off_w;
- p += off_w;
- bd->free += n;
-
- accountAllocation(cap, n);
-
- return p;
}
/* -----------------------------------------------------------------------------
View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/264800330b6e83ec1405ba8f646a36c3cdcbc300
--
View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/264800330b6e83ec1405ba8f646a36c3cdcbc300
You're receiving this email because of your account on gitlab.haskell.org.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.haskell.org/pipermail/ghc-commits/attachments/20230524/94598f85/attachment-0001.html>
More information about the ghc-commits
mailing list