As a follow-up to the previous commit, move the recovery related code in
__gfs2_glock_dq() to gfs2_glock_dq() where it better fits. No
functional change.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Remove the support for glock holder auto-demotion (commit dc732906c2
and folow-ups) as we are not planning to use this feature, and the
additional code therefore only adds unnecessary complexity.
Signed-off-by: Andreas Gruenbacher <[email protected]>
In gfs2_try_evict(), when an inode can't be evicted, we are grabbing a
temporary reference on the inode glock to poke that glock. That should
be safe, but it's easier to just grab an inode reference as we already
do earlier in this function.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Commit c412a97cf6 changed delete_work_func() to always perform an
inode lookup when gfs2_try_evict() fails. This doesn't make sense as a
gfs2_try_evict() failure indicates that the inode is likely still in
use. Revert that change.
Fixes: c412a97cf6 ("gfs2: Use TRY lock in gfs2_inode_lookup for UNLINKED inodes")
Signed-off-by: Andreas Gruenbacher <[email protected]>
Add comment on when and why gfs2_cancel_delete_work() needs to be
skipped in gfs2_inode_lookup().
Signed-off-by: Andreas Gruenbacher <[email protected]>
Those functions have reached a size at which having them inline isn't
useful anymore, so uninline them. In addition, report the glock name on
assertion failures.
Signed-off-by: Andreas Gruenbacher <[email protected]>
With the previous change, to simplify things, we can always just dequeue
and uninitialize the iopen glock in gfs2_evict_inode() even if it isn't
queued anymore.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Since commit 3d36e57ff7 ("gfs2: gfs2_create_inode rework"),
gfs2_evict_inode() and gfs2_create_inode() / gfs2_inode_lookup() will
synchronize via the inode hash table and we can be certain that once a
new inode is inserted into the inode hash table(), gfs2_evict_inode()
has completely destroyed any previous versions. We no longer need to
worry about overlapping inode object lifespans. Update the code and
comments accordingly.
Signed-off-by: Andreas Gruenbacher <[email protected]>
When a locking request fails, the associated glock holder is
automatically dequeued from the list of active and waiting holders. For
GL_ASYNC locking requests, this will obviously happen asynchronously
and it can race with attempts to cancel that locking request via
gfs2_glock_dq(). Therefore, don't forget to check if a locking request
has already been dequeued in gfs2_glock_dq().
Signed-off-by: Andreas Gruenbacher <[email protected]>
Check if the inode size of stuffed (inline) inodes is within the allowed
range when reading inodes from disk (gfs2_dinode_in()). This prevents
us from on-disk corruption.
The two checks in stuffed_readpage() and gfs2_unstuffer_page() that just
truncate inline data to the maximum allowed size don't actually make
sense, and they can be removed now as well.
Reported-by: [email protected]
Signed-off-by: Andreas Gruenbacher <[email protected]>
In each of the two functions, add an inode variable that points to
&ip->i_inode and use that throughout the rest of the function.
Signed-off-by: Andreas Gruenbacher <[email protected]>
When creating a new inode, there is a small chance that an inode lookup
for a previous version of the same inode is still in progress. In that
case, that previous lookup will eventually fail, but we may still need
to retry here.
Signed-off-by: Andreas Gruenbacher <[email protected]>
When gfs2_create_inode() fails after creating a new inode, it uses the
GIF_FREE_VFS_INODE and GIF_ALLOC_FAILED inode flags to communicate to
gfs2_evict_inode() which parts of the inode need to be deallocated and
destroyed. In some error cases, the inode ends up being allocated on
disk and then accidentally left behind. In others, the inode is
partially constructed and then not properly destroyed. Clean this up by
completely handling the inode deallocation and destruction in
gfs2_evict_inode().
This means that gfs2_evict_inode() may now be faced with partially
constructed inodes, so add the necessary checks to cope with that. In
particular, make sure that for incompletely constructed inodes, we're
not accessing the buffers backing the on-disk blocks; the contents may
be undefined.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Initialize variable "ip" earlier so that it can be used interchangeably
with "inode" everywhere.
Signed-off-by: Andreas Gruenbacher <[email protected]>
In gfs2_create_inode, get rid of the ghs array in favor of two separate
variables. This makes the code much less irritating.
Signed-off-by: Andreas Gruenbacher <[email protected]>
We have reserved the number of blocks we want to allocate, so the actual
allocation isn't expected to fail. Nevertheless, make the code behave
correctly even when things go wrong.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Resolves a conflict in gfs2_inode_lookup() between the following commits:
gfs2: Use TRY lock in gfs2_inode_lookup for UNLINKED inodes
gfs2: Mark the remaining process-independent glock holders as GL_NOPID
Signed-off-by: Andreas Gruenbacher <[email protected]>
Switch from strlcpy to strscpy and make sure that @count is the size of
the smaller of the source and destination buffers. This prevents
reading beyond the end of the source buffer when the source string isn't
null terminated.
Found by a modified version of syzkaller.
Suggested-by: Wolfram Sang <[email protected]>
Signed-off-by: Andreas Gruenbacher <[email protected]>
In do_promote(), we're never removing the current entry from the list
and so the list traversal is actually safe. Switch back to
list_for_each_entry().
Signed-off-by: Andreas Gruenbacher <[email protected]>
In do_promote() and add_to_queue(), use current_gh as the variable name
for the first strong holder we could find: this matches the variable
name is may_grant(), and more clearly indicates that we're interested in
one (any) of the current strong holders.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Make go_instantiate take a glock instead of a glock holder as its argument:
this handler is supposed to instantiate the object associated with the glock.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Right now, inode_go_instantiate() contains functionality that relates to
how a glock is held rather than the glock itself, like waiting for
pending direct I/O to complete and completing interrupted truncates.
This code is meant to be run each time a holder is acquired, but
go_instantiate is actually only called once, when the glock is
instantiated.
To fix that, introduce a new go_held glock operation that is called each
time a glock holder is acquired. Move the holder specific code in
inode_go_instantiate() over to inode_go_held().
Signed-off-by: Andreas Gruenbacher <[email protected]>
Now that interrupted truncates are completed in the context of the
process taking the glock, there is no need for the glock state engine to
delegate that task to gfs2_quotad or for quotad to perform those
truncates anymore. Get rid of the obsolete associated infrastructure.
Reverts commit 813e0c46c9 ("GFS2: Fix "truncate in progress" hang").
Signed-off-by: Andreas Gruenbacher <[email protected]>
Signed-off-by: Bob Peterson <[email protected]>
Instantiate glocks outside of the glock state engine: there is no real
reason for instantiating them inside the glock state engine; it only
complicates the code.
Instead, instantiate them in gfs2_glock_wait() and gfs2_glock_async_wait()
using the new gfs2_glock_holder_ready() helper. On top of that, the only
other place that acquires a glock without using gfs2_glock_wait() or
gfs2_glock_async_wait() is gfs2_upgrade_iopen_glock(), so call
gfs2_glock_holder_ready() there as well.
If a dinode has a pending truncate, the glock-specific instantiate function
for inodes wakes up the truncate function in the quota daemon. Waiting for
the completion of the truncate was previously done by the glock state
engine, but we now need to wait in inode_go_instantiate().
This also means that gfs2_instantiate() will now no longer return any
"special" error codes.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Since commit 1fc05c8d84 ("gfs2: cancel timed-out glock requests"), a
pending locking request can be canceled by calling gfs2_glock_dq() on
the pending holder. In gfs2_glock_async_wait(), when we time out, use
that to cancel the remaining locking requests and dequeue the locking
requests already granted. That's simpler as well as more efficient than
waiting for all locking requests to eventually be granted and dequeuing
them then.
In addition, gfs2_glock_async_wait() promises that by the time the
function completes, all glocks are either granted or dequeued, but the
implementation doesn't keep that promise if individual locking requests
fail. Fix that as well.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Add the GL_NOPID flag for the remaining glock holders which are not
associated with the current process.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Add a GL_NOPID flag to indicate that once a glock holder has been acquired, it
won't be associated with the current process anymore. This is useful for iopen
and flock glocks which are associated with open files, as well as journal glock
holders and similar which are associated with the filesystem.
Once GL_NOPID is used for all applicable glocks (see the next patches),
processes will no longer be falsely reported as holding glocks which they are
not actually holding in the glocks dump file. Unlike before, when a process is
reported as having "(ended)", this will indicate an actual bug.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Include flock glocks in the "glockfd" debugfs file. Those are similar to the
iopen glocks; while an open file is holding an flock, it is holding the file's
flock glock.
We cannot take f_fl_mutex in gfs2_glockfd_seq_show_flock() or else dumping the
"glockfd" file would block on flock operations. Instead, use the file->f_lock
spin lock to protect the f_fl_gh.gh_gl glock pointer.
Signed-off-by: Andreas Gruenbacher <[email protected]>
When a process has a gfs2 file open, the file is keeping a reference on the
underlying gfs2 inode, and the inode is keeping the inode's iopen glock held in
shared mode. In other words, the process depends on the iopen glock of each
open gfs2 file. Expose those dependencies in a new "glockfd" debugfs file.
The new debugfs file contains one line for each gfs2 file descriptor,
specifying the tgid, file descriptor number, and glock name, e.g.,
1601 6 5/816d
This list is compiled by iterating all tasks on the system using find_ge_pid(),
and all file descriptors of each task using task_lookup_next_fd_rcu(). To make
that work from gfs2, export those two functions.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Add state and flags arguments to gfs2_rlist_alloc() to make it somewhat more
obvious which state and flags an rlist uses. With that, stop knocking off
flags in gfs2_glock_nq_m() and its nq_m_sync() helper that are never set in the
first place.
Signed-off-by: Andreas Gruenbacher <[email protected]>
We're having unresolved issues with the glock holder auto-demotion mechanism
introduced in commit dc732906c2. This mechanism was assumed to be essential
for avoiding frequent short reads and writes until commit 296abc0d91
("gfs2: No short reads or writes upon glock contention"). Since then,
when the inode glock is lost, it is simply re-acquired and the operation
is resumed. This means that apart from the performance penalty, we
might as well drop the inode glock before faulting in pages, and
re-acquire it afterwards.
Signed-off-by: Andreas Gruenbacher <[email protected]>
In gfs2_file_buffered_write, to increase the likelihood that all the
user memory we're trying to write will be resident in memory, carry out
the write in chunks and fault in each chunk of user memory before trying
to write it. Otherwise, some workloads will trigger frequent short
"internal" writes, causing filesystem blocks to be allocated and then
partially deallocated again when writing into holes, which is wasteful
and breaks reservations.
Neither the chunked writes nor any of the short "internal" writes are
user visible.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Align the chunks that reads and writes are carried out in to the page
cache rather than the user buffers. This will be more efficient in
general, especially for allocating writes. Optimizing the case that the
user buffer is gfs2 backed isn't very useful; we only need to make sure
we won't deadlock.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Pull the return value test of the previous read or write operation out
of should_fault_in_pages(). In a following patch, we'll fault in pages
before the I/O and there will be no return value to check.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Instead of counting the number of bytes read from the filesystem,
functions gfs2_file_direct_read and gfs2_file_read_iter count the number
of bytes written into the user buffer. Conversely, functions
gfs2_file_direct_write and gfs2_file_buffered_write count the number of
bytes read from the user buffer. This is nothing but confusing, so
change the read functions to count how many bytes they have read, and
the write functions to count how many bytes they have written.
Signed-off-by: Andreas Gruenbacher <[email protected]>
When a write cannot be carried out in full, gfs2_iomap_end() releases
blocks that have been allocated for this write but haven't been used.
To compute the end of the allocation, gfs2_iomap_end() incorrectly
rounded the end of the attempted write down to the next block boundary
to arrive at the end of the allocation. It would have to round up, but
the end of the allocation is also available as iomap->offset +
iomap->length, so just use that instead.
In addition, use round_up() for computing the start of the unused range.
Fixes: 64bc06bb32 ("gfs2: iomap buffered write support")
Signed-off-by: Andreas Gruenbacher <[email protected]>
In iomap_write_end(), only call iomap_write_failed() on the byte range
that has failed. This should improve code readability, but doesn't fix
an actual bug because iomap_write_failed() is called after updating the
file size here and it only affects the memory beyond the end of the
file.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Reviewed-by: Darrick J. Wong <[email protected]>
Signed-off-by: Darrick J. Wong <[email protected]>
The @lend parameter of truncate_pagecache_range() should be the offset
of the last byte of the hole, not the first byte beyond it.
Fixes: ae259a9c85 ("fs: introduce iomap infrastructure")
Signed-off-by: Andreas Gruenbacher <[email protected]>
Reviewed-by: Darrick J. Wong <[email protected]>
Signed-off-by: Darrick J. Wong <[email protected]>
Commit 00bfe02f47 ("gfs2: Fix mmap + page fault deadlocks for buffered
I/O") changed gfs2_file_read_iter() and gfs2_file_buffered_write() to
allow dropping the inode glock while faulting in user buffers. When the
lock was dropped, a short result was returned to indicate that the
operation was interrupted.
As pointed out by Linus (see the link below), this behavior is broken
and the operations should always re-acquire the inode glock and resume
the operation instead.
Link: https://lore.kernel.org/lkml/CAHk-=whaz-g_nOOoo8RRiWNjnv2R+h6_xk2F1J4TuSRxk1MtLw@mail.gmail.com/
Fixes: 00bfe02f47 ("gfs2: Fix mmap + page fault deadlocks for buffered I/O")
Signed-off-by: Andreas Gruenbacher <[email protected]>
Only re-check for direct I/O writes past the end of the file after
re-acquiring the inode glock.
Signed-off-by: Andreas Gruenbacher <[email protected]>
When part of the user buffer passed to generic_perform_write() or
iomap_file_buffered_write() cannot be faulted in for reading, the entire
write currently fails. The correct behavior would be to write all the
data that can be written, up to the point of failure.
Commit a6294593e8 ("iov_iter: Turn iov_iter_fault_in_readable into
fault_in_iov_iter_readable") gave us the information needed, so fix the
page prefaulting in generic_perform_write() and iomap_write_iter() to
only bail out when no pages could be faulted in.
We already factor in that pages that are faulted in may no longer be
resident by the time they are accessed. Paging out pages has the same
effect as not faulting in those pages in the first place, so the code
can already deal with that.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Reviewed-by: Catalin Marinas <[email protected]>
Reviewed-by: Christoph Hellwig <[email protected]>
When direct writes fail with -ENOTBLK because we're writing into a
hole (gfs2_iomap_begin()) or because of a page invalidation failure
(iomap_dio_rw()), we're falling back to buffered writes. In that case,
when we lose the inode glock in gfs2_file_buffered_write(), we want to
re-acquire it instead of returning a short write.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Function iomap_dio_rw() only returns -ENOTBLK for write requests and
gfs2_file_direct_read() no longer returns -ENOTBLK since commit
1d45bb7f9d ("gfs2: Use iomap for stuffed direct I/O reads"), so there
is no need to check for -ENOTBLK in gfs2_file_read_iter() anymore.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Since commit 554c577cee, gfs2_file_buffered_write() can accidentally
return a truncated iov_iter, which might confuse callers. Fix that.
Fixes: 554c577cee ("gfs2: Prevent endless loops in gfs2_file_buffered_write")
Signed-off-by: Andreas Gruenbacher <[email protected]>
During lockless buffered reads, filemap_read() holds page cache page
references while trying to copy data to the user-space buffer. The
calling process isn't holding the inode glock, but the page references
it holds prevent those pages from being removed from the page cache, and
that prevents the underlying inode glock from being moved to another
node. Thus, we can end up in the same kinds of distributed deadlock
situations as with normal (non-lockless) buffered reads.
Fix that by disabling page faults during lockless reads as well.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Fix the fault-in window size logic:
* Use a maximum window size of 1 MiB instead of BIO_MAX_VECS * PAGE_SIZE.
The previous window size was always one page because the pages variable
was accidentally being defined and then redefined in
should_fault_in_pages().
* The nr_dirtied heuristic for guessing when there might be memory
pressure often results in very small window sizes. Don't let
nr_dirtied drop below 8 pages (as btrfs does).
* Compute the window size in units of bytes, not pages.
* Account for page overlap (unaligned iterators).
Signed-off-by: Andreas Gruenbacher <[email protected]>
The gh_error field if a glock holder is initialized to zero in
gfs2_holder_init(). When a locking operation fails, gh_error is set to
an error code; when it succeeds, the gh_error value is left unchanged.
The field isn't initialized in gfs2_holder_reinit(), which is a problem.
Instead of fixing that directly, initialize gh_error in gfs2_glock_nq().
That also obsoletes the assignment in do_flock().
Signed-off-by: Andreas Gruenbacher <[email protected]>
This patch tries to fix the continual ABBA deadlocks we keep having
between the iopen and inode glocks. This switches the lock order in
gfs2_inode_lookup and gfs2_create_inode so the iopen glock is always
locked first.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Signed-off-by: Bob Peterson <[email protected]>
The gfs2 evict code tries to upgrade the iopen glock from SH to EX. If
the attempt to upgrade times out, gfs2 needs to tell dlm to cancel the
lock request or it can deadlock. We also need to wake up the process
waiting for the lock when dlm sends its AST back to gfs2.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Signed-off-by: Bob Peterson <[email protected]>
Due to the asynchronous nature of the dlm api, when we request a pending
locking request to be canceled with dlm_unlock(DLM_LKF_CANCEL), the
locking request will either complete before it could be canceled, or the
cancellation will succeed. In either case, gdlm_ast will be called once
and the status will indicate the outcome of the locking request, with
-DLM_ECANCEL indicating a canceled request.
Inside dlm, when a locking request completes before its cancel request
could be processed, gdlm_ast will be called, but the lock will still be
considered busy until a DLM_MSG_CANCEL_REPLY message completes the
cancel request. During that time, successive dlm_lock() or dlm_unlock()
requests for that lock will return -EBUSY. In other words, waiting for
the gdlm_ast call before issuing the next locking request is not enough.
There is no way of waiting for a cancel request to actually complete,
either.
We rarely cancel locking requests, but when we do, we don't know when
the next locking request for that lock will occur. This means that any
dlm_lock() or dlm_unlock() call can potentially return -EBUSY. When
that happens, this patch simply repeats the request after a short pause.
This workaround could be improved upon by tracking for which dlm locks
cancel requests have been issued, but that isn't strictly necessary and
it would complicate the code. We haven't seen -EBUSY errors from dlm
without cancel requests.
Signed-off-by: Andreas Gruenbacher <[email protected]>
When gfs2_setattr_size() fails, it calls gfs2_rs_delete(ip, NULL) to get
rid of any reservations the inode may have. Instead, it should pass in
the inode's write count as the second parameter to allow
gfs2_rs_delete() to figure out if the inode has any writers left.
In a next step, there are two instances of gfs2_rs_delete(ip, NULL) left
where we know that there can be no other users of the inode. Replace
those with gfs2_rs_deltree(&ip->i_res) to avoid the unnecessary write
count check.
With that, gfs2_rs_delete() is only called with the inode's actual write
count, so get rid of the second parameter.
Fixes: a097dc7e24 ("GFS2: Make rgrp reservations part of the gfs2_inode structure")
Signed-off-by: Andreas Gruenbacher <[email protected]>
It turns out that the might_sleep() call that commit 660a6126f8 adds
is triggering occasional data corruption in testing. We're not sure
about the root cause yet, but since this commit was added as a debugging
aid only, revert it for now.
This reverts commit 660a6126f8.
Fixes: 660a6126f8 ("gfs2: check context in gfs2_glock_put")
Cc: [email protected] # v5.16+
Signed-off-by: Andreas Gruenbacher <[email protected]>
The description of gfs2_instantiate accidentally lists a glock argument,
but the function takes a glock holder.
Signed-off-by: Andreas Gruenbacher <[email protected]>
The function name in the kernel-doc comment wasn't updated when the
function was renamed.
Fixes: b016d9a84a ("gfs2: Save ip from gfs2_glock_nq_init")
Signed-off-by: Andreas Gruenbacher <[email protected]>
When gfs2_lookup_by_inum() calls gfs2_inode_lookup() for an uncached
inode, gfs2_inode_lookup() will place a new tentative inode into the
inode cache before verifying that there is a valid inode at the given
address. This can race with gfs2_create_inode() which doesn't check for
duplicates inodes. gfs2_create_inode() will try to assign the new inode
to the corresponding inode glock, and glock_set_object() will complain
that the glock is still in use by gfs2_inode_lookup's tentative inode.
We noticed this bug after adding commit 486408d690 ("gfs2: Cancel
remote delete work asynchronously") which allowed delete_work_func() to
race with gfs2_create_inode(), but the same race exists for
open-by-handle.
Fix that by switching from insert_inode_hash() to
insert_inode_locked4(), which does check for duplicate inodes. We know
we've just managed to to allocate the new inode, so an inode tentatively
created by gfs2_inode_lookup() will eventually go away and
insert_inode_locked4() will always succeed.
In addition, don't flush the inode glock work anymore (this can now only
make things worse) and clean up glock_{set,clear}_object for the inode
glock somewhat.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Rework gfs2_inode_lookup() to only set up the new inode's glocks after
verifying that the new inode is valid.
There is no need for flushing the inode glock work queue anymore now,
so remove that as well.
While at it, get rid of the useless wrapper around iget5_locked() and
its unnecessary is_bad_inode() check.
Signed-off-by: Andreas Gruenbacher <[email protected]>
In gfs2_inode_lookup, once the inode has been looked up, we check if the
inode generation (no_formal_ino) is the one we're looking for. If it
isn't and the inode wasn't in the inode cache, we discard the newly
looked up inode. This is unnecessary, complicates the code, and makes
future changes to gfs2_inode_lookup harder, so change the code to retain
newly looked up inodes instead.
Signed-off-by: Andreas Gruenbacher <[email protected]>
When we mock up a temporary holder in gfs2_glock_cb to demote weak holders in
response to a remote locking conflict, we don't set the HIF_HOLDER flag. This
causes function may_grant to BUG. Fix by setting the missing HIF_HOLDER flag
in the mock glock holder.
In addition, define the mock glock holder where it is used.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Change iomap_read_inline_data to return 0 or an error code; this
simplifies the callers. Add a description.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Reviewed-by: Christoph Hellwig <[email protected]>
[djwong: document the return value of iomap_read_inline_data explicitly]
Reviewed-by: Darrick J. Wong <[email protected]>
Signed-off-by: Darrick J. Wong <[email protected]>
Before commit 740499c784 ("iomap: fix the iomap_readpage_actor return
value for inline data"), when hitting an IOMAP_INLINE extent,
iomap_readpage_actor would report having read the entire page. Since
then, it only reports having read the inline data (iomap->length).
This will force iomap_readpage into another iteration, and the
filesystem will report an unaligned hole after the IOMAP_INLINE extent.
But iomap_readpage_actor (now iomap_readpage_iter) isn't prepared to
deal with unaligned extents, it will get things wrong on filesystems
with a block size smaller than the page size, and we'll eventually run
into the following warning in iomap_iter_advance:
WARN_ON_ONCE(iter->processed > iomap_length(iter));
Fix that by changing iomap_readpage_iter to return 0 when hitting an
inline extent; this will cause iomap_iter to stop immediately.
To fix readahead as well, change iomap_readahead_iter to pass on
iomap_readpage_iter return values less than or equal to zero.
Fixes: 740499c784 ("iomap: fix the iomap_readpage_actor return value for inline data")
Cc: [email protected] # v5.15+
Signed-off-by: Andreas Gruenbacher <[email protected]>
Reviewed-by: Christoph Hellwig <[email protected]>
Reviewed-by: Darrick J. Wong <[email protected]>
Signed-off-by: Darrick J. Wong <[email protected]>
Currently, instead of performing a short write,
iomap_file_buffered_write will fail when part of its iov iterator cannot
be read. In contrast, gfs2_file_buffered_write will loop around if it
can read part of the iov iterator, so we can end up in an endless loop.
This should be fixed in iomap_file_buffered_write (and also
generic_perform_write), but this comes a bit late in the 5.16
development cycle, so work around it in the filesystem by
trimming the iov iterator to the known-good size for now.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Function demote_incompat_holders iterates over the list of glock holders
with list_for_each_entry, and it then sometimes removes the current
holder from the list. This will get the loop stuck; we must use
list_for_each_entry_safe instead.
Reported-by: Dan Carpenter <[email protected]>
Signed-off-by: Andreas Gruenbacher <[email protected]>
Fix the length of holes reported at the end of a file: the length is
relative to the beginning of the extent, not the seek position which is
rounded down to the filesystem block size.
This bug went unnoticed for some time, but is now caught by the
following assertion in iomap_iter_done():
WARN_ON_ONCE(iter->iomap.offset + iter->iomap.length <= iter->pos)
Signed-off-by: Andreas Gruenbacher <[email protected]>
Replace test_bit() + set_bit() with test_and_set_bit() where we need an atomic
operation. Use clear_and_wake_up_bit() instead of open coding it.
Signed-off-by: Andreas Gruenbacher <[email protected]>
So far, glock_hash_walk took a reference on each glock it iterated over, and it
was the examiner's responsibility to drop those references. Dropping the final
reference to a glock can sleep and the examiners are called in a RCU critical
section with spin locks held, so examiners that didn't need the extra reference
had to drop it asynchronously via gfs2_glock_queue_put or similar. This wasn't
done correctly in thaw_glock which did call gfs2_glock_put, and not at all in
dump_glock_func.
Change glock_hash_walk to not take glock references at all. That way, the
examiners that don't need them won't have to bother with slow asynchronous
puts, and the examiners that do need references can take them themselves.
Reported-by: Alexander Aring <[email protected]>
Signed-off-by: Andreas Gruenbacher <[email protected]>
In gfs2_inode_lookup and gfs2_create_inode, we're calling
gfs2_cancel_delete_work which currently cancels any remote delete work
(delete_work_func) synchronously. This means that if the work is
currently running, it will wait for it to finish. We're doing this to
pevent a previous instance of an inode from having any influence on the
next instance.
However, delete_work_func uses gfs2_inode_lookup internally, and we can
end up in a deadlock when delete_work_func gets interrupted at the wrong
time. For example,
(1) An inode's iopen glock has delete work queued, but the inode
itself has been evicted from the inode cache.
(2) The delete work is preempted before reaching gfs2_inode_lookup.
(3) Another process recreates the inode (gfs2_create_inode). It tries
to cancel any outstanding delete work, which blocks waiting for
the ongoing delete work to finish.
(4) The delete work calls gfs2_inode_lookup, which blocks waiting for
gfs2_create_inode to instantiate and unlock the new inode =>
deadlock.
It turns out that when the delete work notices that its inode has been
re-instantiated, it will do nothing. This means that it's safe to
cancel the delete work asynchronously. This prevents the kind of
deadlock described above.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Signed-off-by: Bob Peterson <[email protected]>
Remove the 'first' argument of trace_gfs2_promote: with GL_SKIP, the
'first' holder isn't the one that instantiates the glock
(gl_instantiate), which is what the 'first' flag was apparently supposed
to indicate.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Before this patch, when a glock was locked by function gfs2_glock_nq_init,
it initialized the holder gh_ip (return address) as gfs2_glock_nq_init.
That made it extremely difficult to track down problems because many
functions call gfs2_glock_nq_init. This patch changes the function so
that it saves gh_ip from the caller of gfs2_glock_nq_init, which makes
it easy to backtrack which holder took the lock.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Signed-off-by: Bob Peterson <[email protected]>
Also disable page faults during direct I/O requests and implement a
similar kind of retry logic as in the buffered I/O case.
The retry logic in the direct I/O case differs from the buffered I/O
case in the following way: direct I/O doesn't provide the kinds of
consistency guarantees between concurrent reads and writes that buffered
I/O provides, so once we lose the inode glock while faulting in user
pages, we always resume the operation. We never need to return a
partial read or write.
This locking problem was originally reported by Jan Kara. Linus came up
with the idea of disabling page faults. Many thanks to Al Viro and
Matthew Wilcox for their feedback.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Introduce a new nofault flag to indicate to iov_iter_get_pages not to
fault in user pages.
This is implemented by passing the FOLL_NOFAULT flag to get_user_pages,
which causes get_user_pages to fail when it would otherwise fault in a
page. We'll use the ->nofault flag to prevent iomap_dio_rw from faulting
in pages when page faults are not allowed.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Introduce a new FOLL_NOFAULT flag that causes get_user_pages to return
-EFAULT when it would otherwise trigger a page fault. This is roughly
similar to FOLL_FAST_ONLY but available on all architectures, and less
fragile.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Add a done_before argument to iomap_dio_rw that indicates how much of
the request has already been transferred. When the request succeeds, we
report that done_before additional bytes were tranferred. This is
useful for finishing a request asynchronously when part of the request
has already been completed synchronously.
We'll use that to allow iomap_dio_rw to be used with page faults
disabled: when a page fault occurs while submitting a request, we
synchronously complete the part of the request that has already been
submitted. The caller can then take care of the page fault and call
iomap_dio_rw again for the rest of the request, passing in the number of
bytes already tranferred.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Reviewed-by: Darrick J. Wong <[email protected]>
In iomap_dio_rw, when iomap_apply returns an -EFAULT error and the
IOMAP_DIO_PARTIAL flag is set, complete the request synchronously and
return a partial result. This allows the caller to deal with the page
fault and retry the remainder of the request.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Reviewed-by: Darrick J. Wong <[email protected]>
When a user copy fails in one of the helpers of iomap_dio_rw, fail with
-EFAULT instead of returning 0. This matches what iomap_dio_bio_actor
returns when it gets an -EFAULT from bio_iov_iter_get_pages. With these
changes, iomap_dio_actor now consistently fails with -EFAULT when a user
page cannot be faulted in.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Reviewed-by: Darrick J. Wong <[email protected]>
Reviewed-by: Christoph Hellwig <[email protected]>
In the .read_iter and .write_iter file operations, we're accessing
user-space memory while holding the inode glock. There is a possibility
that the memory is mapped to the same file, in which case we'd recurse
on the same glock.
We could detect and work around this simple case of recursive locking,
but more complex scenarios exist that involve multiple glocks,
processes, and cluster nodes, and working around all of those cases
isn't practical or even possible.
Avoid these kinds of problems by disabling page faults while holding the
inode glock. If a page fault would occur, we either end up with a
partial read or write or with -EFAULT if nothing could be read or
written. In either case, we know that we're not done with the
operation, so we indicate that we're willing to give up the inode glock
and then we fault in the missing pages. If that made us lose the inode
glock, we return a partial read or write. Otherwise, we resume the
operation.
This locking problem was originally reported by Jan Kara. Linus came up
with the idea of disabling page faults. Many thanks to Al Viro and
Matthew Wilcox for their feedback.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Now that gfs2_file_buffered_write is the only remaining user of
ip->i_gh, we can move the glock holder to the stack (or rather, use the
one we already have on the stack); there is no need for keeping the
holder in the inode anymore.
This is slightly complicated by the fact that we're using ip->i_gh for
the statfs inode in gfs2_file_buffered_write as well. Writing to the
statfs inode isn't very common, so allocate the statfs holder
dynamically when needed.
Signed-off-by: Andreas Gruenbacher <[email protected]>
So far, for buffered writes, we were taking the inode glock in
gfs2_iomap_begin and dropping it in gfs2_iomap_end with the intention of
not holding the inode glock while iomap_write_actor faults in user
pages. It turns out that iomap_write_actor is called inside iomap_begin
... iomap_end, so the user pages were still faulted in while holding the
inode glock and the locking code in iomap_begin / iomap_end was
completely pointless.
Move the locking into gfs2_file_buffered_write instead. We'll take care
of the potential deadlocks due to faulting in user pages while holding a
glock in a subsequent patch.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Pass the first current glock holder into function may_grant and
deobfuscate the logic there.
While at it, switch from BUG_ON to GLOCK_BUG_ON in may_grant. To make
that build cleanly, de-constify the may_grant arguments.
We're now using function find_first_holder in do_promote, so move the
function's definition above do_promote.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Add a wrapper around iomap_file_buffered_write. We'll add code for when
the operation needs to be retried here later.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Introduce a new fault_in_iov_iter_writeable helper for safely faulting
in an iterator for writing. Uses get_user_pages() to fault in the pages
without actually writing to them, which would be destructive.
We'll use fault_in_iov_iter_writeable in gfs2 once we've determined that
the iterator passed to .read_iter isn't in memory.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Turn iov_iter_fault_in_readable into a function that returns the number
of bytes not faulted in, similar to copy_to_user, instead of returning a
non-zero value when any of the requested pages couldn't be faulted in.
This supports the existing users that require all pages to be faulted in
as well as new users that are happy if any pages can be faulted in.
Rename iov_iter_fault_in_readable to fault_in_iov_iter_readable to make
sure this change doesn't silently break things.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Turn fault_in_pages_{readable,writeable} into versions that return the
number of bytes not faulted in, similar to copy_to_user, instead of
returning a non-zero value when any of the requested pages couldn't be
faulted in. This supports the existing users that require all pages to
be faulted in as well as new users that are happy if any pages can be
faulted in.
Rename the functions to fault_in_{readable,writeable} to make sure
this change doesn't silently break things.
Neither of these functions is entirely trivial and it doesn't seem
useful to inline them, so move them to mm/gup.c.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Both iov_iter_get_pages and iov_iter_get_pages_alloc return the number
of bytes of the iovec they could get the pages for. When they cannot
get any pages, they're supposed to return 0, but when the start of the
iovec isn't page aligned, the calculation goes wrong and they return a
negative value. Fix both functions.
In addition, change iov_iter_get_pages_alloc to return NULL in that case
to prevent resource leaks.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Reviewed-by: Christoph Hellwig <[email protected]>
The permission check in gfs2_setattr is an old and outdated version of
may_setattr(). Switch to the updated version.
Fixes fstest generic/079.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Signed-off-by: Bob Peterson <[email protected]>
Signed-off-by: Al Viro <[email protected]>
Move the permission checks in notify_change into a separate function to
make them available to filesystems.
When notify_change is called, the vfs performs those checks before
calling into iop->setattr. However, a filesystem like gfs2 can only
lock and revalidate the inode inside ->setattr, and it must then repeat
those checks to err on the safe side.
It would be nice to get rid of the double checking, but moving the
permission check into iop->setattr altogether isn't really an option.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Signed-off-by: Bob Peterson <[email protected]>
Signed-off-by: Al Viro <[email protected]>
Fix some typos and bad grammar in buffered-io.c to make the comments
easier to read.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Reviewed-by: Darrick J. Wong <[email protected]>
Signed-off-by: Darrick J. Wong <[email protected]>
Now that we create those objects in iomap_writepage_map when needed,
there's no need to pre-create them in iomap_page_mkwrite_actor anymore.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Reviewed-by: Christoph Hellwig <[email protected]>
Reviewed-by: Matthew Wilcox (Oracle) <[email protected]>
Reviewed-by: Darrick J. Wong <[email protected]>
Signed-off-by: Darrick J. Wong <[email protected]>
In iomap_readpage_actor, don't create iop objects for inline inodes.
Otherwise, iomap_read_inline_data will set PageUptodate without setting
iop->uptodate, and iomap_page_release will eventually complain.
To prevent this kind of bug from occurring in the future, make sure the
page doesn't have private data attached in iomap_read_inline_data.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Reviewed-by: Christoph Hellwig <[email protected]>
Reviewed-by: Darrick J. Wong <[email protected]>
Signed-off-by: Darrick J. Wong <[email protected]>
Create an iop in the writeback path if one doesn't exist. This allows us
to avoid creating the iop in some cases. We'll initially do that for pages
with inline data, but it can be extended to pages which are entirely within
an extent. It also allows for an iop to be removed from pages in the
future (eg page split).
Co-developed-by: Matthew Wilcox (Oracle) <[email protected]>
Signed-off-by: Matthew Wilcox (Oracle) <[email protected]>
Signed-off-by: Andreas Gruenbacher <[email protected]>
Reviewed-by: Christoph Hellwig <[email protected]>
Reviewed-by: Darrick J. Wong <[email protected]>
Signed-off-by: Darrick J. Wong <[email protected]>
Split __gfs2_unstuff_inode off from gfs2_unstuff_dinode and clean up the
code a little. All remaining callers now pass NULL as the page argument
of gfs2_unstuff_dinode, so remove that argument.
Signed-off-by: Andreas Gruenbacher <[email protected]>
Andreas Gruenbacher