Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Remove some dead ARC code. #14340

Merged
merged 1 commit into from
Jan 9, 2023
Merged

Remove some dead ARC code. #14340

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 0 additions & 1 deletion include/sys/arc.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -195,7 +195,6 @@ typedef enum arc_buf_flags {
struct arc_buf {
arc_buf_hdr_t *b_hdr;
arc_buf_t *b_next;
kmutex_t b_evict_lock;
void *b_data;
arc_buf_flags_t b_flags;
};
Expand Down
100 changes: 26 additions & 74 deletions module/zfs/arc.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1196,7 +1196,6 @@ buf_cons(void *vbuf, void *unused, int kmflag)
arc_buf_t *buf = vbuf;

memset(buf, 0, sizeof (arc_buf_t));
mutex_init(&buf->b_evict_lock, NULL, MUTEX_DEFAULT, NULL);
arc_space_consume(sizeof (arc_buf_t), ARC_SPACE_HDRS);

return (0);
Expand Down Expand Up @@ -1244,9 +1243,8 @@ static void
buf_dest(void *vbuf, void *unused)
{
(void) unused;
arc_buf_t *buf = vbuf;
(void) vbuf;

mutex_destroy(&buf->b_evict_lock);
arc_space_return(sizeof (arc_buf_t), ARC_SPACE_HDRS);
}

Expand Down Expand Up @@ -3506,11 +3504,8 @@ arc_hdr_realloc_crypt(arc_buf_hdr_t *hdr, boolean_t need_crypt)
*/
(void) zfs_refcount_add(&nhdr->b_l1hdr.b_refcnt, FTAG);
nhdr->b_l1hdr.b_buf = hdr->b_l1hdr.b_buf;
for (buf = nhdr->b_l1hdr.b_buf; buf != NULL; buf = buf->b_next) {
mutex_enter(&buf->b_evict_lock);
for (buf = nhdr->b_l1hdr.b_buf; buf != NULL; buf = buf->b_next)
buf->b_hdr = nhdr;
mutex_exit(&buf->b_evict_lock);
}

zfs_refcount_transfer(&nhdr->b_l1hdr.b_refcnt, &hdr->b_l1hdr.b_refcnt);
(void) zfs_refcount_remove(&nhdr->b_l1hdr.b_refcnt, FTAG);
Expand Down Expand Up @@ -3914,11 +3909,13 @@ arc_evict_hdr(arc_buf_hdr_t *hdr, uint64_t *real_evicted)
ASSERT(MUTEX_HELD(HDR_LOCK(hdr)));
ASSERT(HDR_HAS_L1HDR(hdr));
ASSERT(!HDR_IO_IN_PROGRESS(hdr));
ASSERT0(hdr->b_l1hdr.b_bufcnt);
ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL);
ASSERT0(zfs_refcount_count(&hdr->b_l1hdr.b_refcnt));
Comment on lines +3912 to +3914
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

What do you think about making some or all of these VERIFYs to make sure that the assumptions here are valid in non-debug builds?

Copy link
Member Author

@amotin amotin Jan 6, 2023 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is a very hot single-threaded piece of code, limiting system performance on some workloads. I do not want VERIFY() impossible here without a good reason. Any particular reason you think this may work different for debug and non-debug builds?

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

If you're convinced that this can't happen then no need to be defensive. My concern was that if we find that the b_buf is not being destroyed because of some missed code path, then we end up with a memory leak which would be hard to track down. Doing a simple VERIFY3P(hdr->b_l1hdr.b_buf, ==, NULL) would allow for a quick diagnosis on non-debug bits. If you've stressed tested this significantly then I would agree that leaving these off makes sense if they can't happen.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I've added the assertions to protect from some logical bugs. Since it didn't fire yet on any tests, I tend to think it won't, and the assertions will just make sure that whoever modify the code in future will have a guard rail.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

And it likely won't leak any way. It will get evicted to ghost state with the buffer, and on eviction from there arc_hdr_destroy() frees any possible remaining buffers. I thought to remove it also, but appeared it is actually used when arc_buf_destroy() calls remove_reference() for anonymous header without getting the hash lock.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ok. The arc behaves very differently in non-debug builds so some of the inherit race conditions may not show up there. I will trust that you're confident enough in the change.


*real_evicted = 0;
state = hdr->b_l1hdr.b_state;
if (GHOST_STATE(state)) {
ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL);

/*
* l2arc_write_buffers() relies on a header's L1 portion
Expand Down Expand Up @@ -3972,21 +3969,6 @@ arc_evict_hdr(arc_buf_hdr_t *hdr, uint64_t *real_evicted)
return (bytes_evicted);
}

ASSERT0(zfs_refcount_count(&hdr->b_l1hdr.b_refcnt));
while (hdr->b_l1hdr.b_buf) {
arc_buf_t *buf = hdr->b_l1hdr.b_buf;
if (!mutex_tryenter(&buf->b_evict_lock)) {
ARCSTAT_BUMP(arcstat_mutex_miss);
break;
}
if (buf->b_data != NULL) {
bytes_evicted += HDR_GET_LSIZE(hdr);
*real_evicted += HDR_GET_LSIZE(hdr);
}
mutex_exit(&buf->b_evict_lock);
arc_buf_destroy_impl(buf);
}

if (HDR_HAS_L2HDR(hdr)) {
ARCSTAT_INCR(arcstat_evict_l2_cached, HDR_GET_LSIZE(hdr));
} else {
Expand Down Expand Up @@ -4014,32 +3996,27 @@ arc_evict_hdr(arc_buf_hdr_t *hdr, uint64_t *real_evicted)
}
}

if (hdr->b_l1hdr.b_bufcnt == 0) {
arc_cksum_free(hdr);

bytes_evicted += arc_hdr_size(hdr);
*real_evicted += arc_hdr_size(hdr);
bytes_evicted += arc_hdr_size(hdr);
*real_evicted += arc_hdr_size(hdr);

/*
* If this hdr is being evicted and has a compressed
* buffer then we discard it here before we change states.
* This ensures that the accounting is updated correctly
* in arc_free_data_impl().
*/
if (hdr->b_l1hdr.b_pabd != NULL)
arc_hdr_free_abd(hdr, B_FALSE);
/*
* If this hdr is being evicted and has a compressed buffer then we
* discard it here before we change states. This ensures that the
* accounting is updated correctly in arc_free_data_impl().
*/
if (hdr->b_l1hdr.b_pabd != NULL)
arc_hdr_free_abd(hdr, B_FALSE);

if (HDR_HAS_RABD(hdr))
arc_hdr_free_abd(hdr, B_TRUE);
if (HDR_HAS_RABD(hdr))
arc_hdr_free_abd(hdr, B_TRUE);

arc_change_state(evicted_state, hdr);
DTRACE_PROBE1(arc__evict, arc_buf_hdr_t *, hdr);
if (evicted_state == arc_anon) {
arc_hdr_destroy(hdr);
*real_evicted += HDR_FULL_SIZE;
} else {
ASSERT(HDR_IN_HASH_TABLE(hdr));
}
arc_change_state(evicted_state, hdr);
DTRACE_PROBE1(arc__evict, arc_buf_hdr_t *, hdr);
if (evicted_state == arc_anon) {
arc_hdr_destroy(hdr);
*real_evicted += HDR_FULL_SIZE;
} else {
ASSERT(HDR_IN_HASH_TABLE(hdr));
}

return (bytes_evicted);
Expand Down Expand Up @@ -5631,31 +5608,25 @@ arc_access(arc_buf_hdr_t *hdr, arc_flags_t arc_flags, boolean_t hit)
void
arc_buf_access(arc_buf_t *buf)
{
mutex_enter(&buf->b_evict_lock);
arc_buf_hdr_t *hdr = buf->b_hdr;

/*
* Avoid taking the hash_lock when possible as an optimization.
* The header must be checked again under the hash_lock in order
* to handle the case where it is concurrently being released.
*/
if (hdr->b_l1hdr.b_state == arc_anon || HDR_EMPTY(hdr)) {
mutex_exit(&buf->b_evict_lock);
if (hdr->b_l1hdr.b_state == arc_anon || HDR_EMPTY(hdr))
return;
}

kmutex_t *hash_lock = HDR_LOCK(hdr);
mutex_enter(hash_lock);

if (hdr->b_l1hdr.b_state == arc_anon || HDR_EMPTY(hdr)) {
mutex_exit(hash_lock);
mutex_exit(&buf->b_evict_lock);
ARCSTAT_BUMP(arcstat_access_skip);
return;
}

mutex_exit(&buf->b_evict_lock);

ASSERT(hdr->b_l1hdr.b_state == arc_mru ||
hdr->b_l1hdr.b_state == arc_mfu ||
hdr->b_l1hdr.b_state == arc_uncached);
Expand Down Expand Up @@ -6593,8 +6564,6 @@ arc_release(arc_buf_t *buf, const void *tag)
* But we don't know that information at this level.
*/

mutex_enter(&buf->b_evict_lock);

ASSERT(HDR_HAS_L1HDR(hdr));

/*
Expand All @@ -6603,7 +6572,6 @@ arc_release(arc_buf_t *buf, const void *tag)
* linked into the hash table.
*/
if (hdr->b_l1hdr.b_state == arc_anon) {
mutex_exit(&buf->b_evict_lock);
ASSERT(!HDR_IO_IN_PROGRESS(hdr));
ASSERT(!HDR_IN_HASH_TABLE(hdr));
ASSERT(!HDR_HAS_L2HDR(hdr));
Expand Down Expand Up @@ -6753,10 +6721,6 @@ arc_release(arc_buf_t *buf, const void *tag)

mutex_exit(hash_lock);

/*
* Allocate a new hdr. The new hdr will contain a b_pabd
* buffer which will be freed in arc_write().
*/
nhdr = arc_hdr_alloc(spa, psize, lsize, protected,
compress, hdr->b_complevel, type);
ASSERT3P(nhdr->b_l1hdr.b_buf, ==, NULL);
Expand All @@ -6772,11 +6736,9 @@ arc_release(arc_buf_t *buf, const void *tag)
(void) zfs_refcount_add(&nhdr->b_l1hdr.b_refcnt, tag);
buf->b_hdr = nhdr;

mutex_exit(&buf->b_evict_lock);
(void) zfs_refcount_add_many(&arc_anon->arcs_size,
arc_buf_size(buf), buf);
} else {
mutex_exit(&buf->b_evict_lock);
ASSERT(zfs_refcount_count(&hdr->b_l1hdr.b_refcnt) == 1);
/* protected by hash lock, or hdr is on arc_anon */
ASSERT(!multilist_link_active(&hdr->b_l1hdr.b_arc_node));
Expand All @@ -6797,25 +6759,15 @@ arc_release(arc_buf_t *buf, const void *tag)
int
arc_released(arc_buf_t *buf)
{
int released;

mutex_enter(&buf->b_evict_lock);
released = (buf->b_data != NULL &&
return (buf->b_data != NULL &&
buf->b_hdr->b_l1hdr.b_state == arc_anon);
mutex_exit(&buf->b_evict_lock);
return (released);
}

#ifdef ZFS_DEBUG
int
arc_referenced(arc_buf_t *buf)
{
int referenced;

mutex_enter(&buf->b_evict_lock);
referenced = (zfs_refcount_count(&buf->b_hdr->b_l1hdr.b_refcnt));
mutex_exit(&buf->b_evict_lock);
return (referenced);
return (zfs_refcount_count(&buf->b_hdr->b_l1hdr.b_refcnt));
}
#endif

Expand Down