- 07 May, 2020 12 commits
-
-
Brian Foster authored
The shutdown parameter of xfs_trans_ail_remove() is no longer used. The remaining callers use it for items that legitimately might not be in the AIL or from contexts where AIL state has already been checked. Remove the unnecessary parameter and fix up the callers. Signed-off-by:
Brian Foster <bfoster@redhat.com> Reviewed-by:
Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by:
Christoph Hellwig <hch@lst.de> Reviewed-by:
Allison Collins <allison.henderson@oracle.com> Signed-off-by:
-
Brian Foster authored
Various intent log items call xfs_trans_ail_remove() with a log I/O error shutdown type, but this helper historically checks whether an item is in the AIL before calling xfs_trans_ail_delete(). This means the shutdown check is essentially a no-op for users of xfs_trans_ail_remove(). It is possible that some items might not be AIL resident when the AIL remove attempt occurs, but this should be isolated to cases where the filesystem has already shutdown. For example, this includes abort of the transaction committing the intent and I/O error of the iclog buffer committing the intent to the log. Therefore, update these callsites to use xfs_trans_ail_delete() to provide AIL state validation for the common path of items being released and removed when associated done items commit to the physical log. Signed-off-by:
-
Brian Foster authored
Several callers acquire the lock just prior to the call. Callers that require ->ail_lock for other purposes already check IN_AIL state and thus don't require the additional shutdown check in the helper. Push the lock down into xfs_trans_ail_delete(), open code the instances that still acquire it, and remove the unnecessary ailp parameter. Signed-off-by:
-
Brian Foster authored
The dquot flush handler effectively aborts the dquot flush if the filesystem is already shut down, but doesn't actually shut down if the flush fails. Update xfs_qm_dqflush() to consistently abort the dquot flush and shutdown the fs if the flush fails with an unexpected error. Signed-off-by:
Dave Chinner <dchinner@redhat.com> Reviewed-by:
-
Brian Foster authored
The pre-flush dquot verification in xfs_qm_dqflush() duplicates the read verifier by checking the dquot in the on-disk buffer. Instead, verify the in-core variant before it is flushed to the buffer. Fixes: 7224fa48 ("xfs: add full xfs_dqblk verifier") Signed-off-by:
-
Brian Foster authored
At unmount time, XFS emits an alert for every in-core buffer that might have undergone a write error. In practice this behavior is probably reasonable given that the filesystem is likely short lived once I/O errors begin to occur consistently. Under certain test or otherwise expected error conditions, this can spam the logs and slow down the unmount. Now that we have a ratelimit mechanism specifically for buffer alerts, reuse it for the per-buffer alerts in xfs_wait_buftarg(). Also lift the final repair message out of the loop so it always prints and assert that the metadata error handling code has shut down the fs. Signed-off-by:
-
Brian Foster authored
XFS has some inconsistent log message rate limiting with respect to buffer alerts. The metadata I/O error notification uses the generic ratelimited alert, the buffer push code uses a custom rate limit and the similar quiesce time failure checks are not rate limited at all (when they should be). The custom rate limit defined in the buf item code is specifically crafted for buffer alerts. It is more aggressive than generic rate limiting code because it must accommodate a high frequency of I/O error events in a relative short timeframe. Factor out the custom rate limit state from the buf item code into a per-buftarg rate limit so various alerts are limited based on the target. Define a buffer alert helper function and use it for the buffer alerts that are already ratelimited. Signed-off-by:
-
Brian Foster authored
The buffer write failure flag is intended to control the internal write retry that XFS has historically implemented to help mitigate the severity of transient I/O errors. The flag is set when a buffer is resubmitted from the I/O completion path due to a previous failure. It is checked on subsequent I/O completions to skip the internal retry and fall through to the higher level configurable error handling mechanism. The flag is cleared in the synchronous and delwri submission paths and also checked in various places to log write failure messages. There are a couple minor problems with the current usage of this flag. One is that we issue an internal retry after every submission from xfsaild due to how delwri submission clears the flag. This results in double the expected or configured number of write attempts when under sustained failures. Another more subtle issue is that the flag is never cleared on successful I/O completion. This can cause xfs_wait_buftarg() to suggest that dirty buffers are being thrown away due to the existence of the flag, when the reality is that the flag might still be set because the write succeeded on the retry. Clear the write failure flag on successful I/O completion to address both of these problems. This means that the internal retry attempt occurs once since the last time a buffer write failed and that various other contexts only see the flag set when the immediately previous write attempt has failed. Signed-off-by:
-
Brian Foster authored
The shutdown check in xfs_iflush() duplicates checks down in the buffer code. If the fs is shut down, xfs_trans_read_buf_map() always returns an error and falls into the same error path. Remove the unnecessary check along with the warning in xfs_imap_to_bp() that generates excessive noise in the log if the fs is shut down. Signed-off-by:
-
Brian Foster authored
The inode flush code has several layers of error handling between the inode and cluster flushing code. If the inode flush fails before acquiring the backing buffer, the inode flush is aborted. If the cluster flush fails, the current inode flush is aborted and the cluster buffer is failed to handle the initial inode and any others that might have been attached before the error. Since xfs_iflush() is the only caller of xfs_iflush_cluster(), the error handling between the two can be condensed in the top-level function. If we update xfs_iflush_int() to always fall through to the log item update and attach the item completion handler to the buffer, any errors that occur after the first call to xfs_iflush_int() can be handled with a buffer I/O failure. Lift the error handling from xfs_iflush_cluster() into xfs_iflush() and consolidate with the existing error handling. This also replaces the need to release the buffer because failing the buffer with XBF_ASYNC drops the current reference. Signed-off-by:
-
Brian Foster authored
We use the same buffer I/O failure code in a few different places. It's not much code, but it's not necessarily self-explanatory. Factor it into a helper and document it in one place. Signed-off-by:
-
Brian Foster authored
Flush locked log items whose underlying buffers fail metadata writeback are tagged with a special flag to indicate that the flush lock is already held. This is currently implemented in the type specific ->iop_push() callback, but the processing required for such items is not type specific because we're only doing basic state management on the underlying buffer. Factor the failed log item handling out of the inode and dquot ->iop_push() callbacks and open code the buffer resubmit helper into a single helper called from xfsaild_push_item(). This provides a generic mechanism for handling failed metadata buffer writeback with a bit less code. Signed-off-by:
-
- 06 May, 2020 1 commit
-
-
Darrick J. Wong authored
Make sure we release resources properly if we cannot clean out the COW extents in preparation for an extent swap. Fixes: 96987eea ("xfs: cancel COW blocks before swapext") Signed-off-by:
-
- 04 May, 2020 27 commits
-
-
Ira Weiny authored
The functionality in xfs_diflags_to_linux() and xfs_diflags_to_iflags() are nearly identical. The only difference is that *_to_linux() is called after inode setup and disallows changing the DAX flag. Combining them can be done with a flag which indicates if this is the initial setup to allow the DAX flag to be properly set only at init time. So remove xfs_diflags_to_linux() and call the modified xfs_diflags_to_iflags() directly. While we are here simplify xfs_diflags_to_iflags() to take struct xfs_inode and use xfs_ip2xflags() to ensure future diflags are included correctly. Reviewed-by:
Ira Weiny <ira.weiny@intel.com> Signed-off-by:
-
Ira Weiny authored
xfs_inode_supports_dax() should reflect if the inode can support DAX not that it is enabled for DAX. Change the use of xfs_inode_supports_dax() to reflect only if the inode and underlying storage support dax. Add a new function xfs_inode_should_enable_dax() which reflects if the inode should be enabled for DAX. Reviewed-by:
-
Ira Weiny authored
As agreed upon[1]. We make the dax mount option a tri-state. '-o dax' continues to operate the same. We add 'always', 'never', and 'inode' (default). [1] https://lore.kernel.org/lkml/20200405061945.GA94792@iweiny-DESK2.sc.intel.com/Signed-off-by:
-
Ira Weiny authored
In prep for the new tri-state mount option which then introduces XFS_MOUNT_DAX_NEVER. Reviewed-by:
-
Ira Weiny authored
An earlier call of xfs_reinit_inode() from xfs_iget_cache_hit() already handles initialization of i_rwsem. Doing so again is unneeded. Reviewed-by:
-
Christoph Hellwig authored
Signed-off-by:
-
Christoph Hellwig authored
Given how XFS is all based around btrees it doesn't make much sense to offer a totally generic state when we can just use the btree cursor. Signed-off-by:
-
Christoph Hellwig authored
All defer op instance place their own extension of the log item into the dfp_done field. Replace that with a xfs_log_item to improve type safety and make the code easier to follow. Also use the opportunity to improve the ->finish_item calling conventions to place the done log item as the higher level structure before the list_entry used for the individual items. Signed-off-by:
-
Christoph Hellwig authored
Split out a helper that operates on a single xfs_defer_pending structure to untangle the code. Signed-off-by:
-
Christoph Hellwig authored
All defer op instance place their own extension of the log item into the dfp_intent field. Replace that with a xfs_log_item to improve type safety and make the code easier to follow. Signed-off-by:
-
Christoph Hellwig authored
This avoids a per-item indirect call, and also simplifies the interface a bit. Signed-off-by:
-
Christoph Hellwig authored
These are aways called together, and my merging them we reduce the amount of indirect calls, improve type safety and in general clean up the code a bit. Signed-off-by:
-
Christoph Hellwig authored
Create a helper that encapsulates the whole logic to create a defer intent. This reorders some of the work that was done, but none of that has an affect on the operation as only fields that don't directly interact are affected. Signed-off-by:
-
Christoph Hellwig authored
Signed-off-by:
-
Christoph Hellwig authored
Signed-off-by:
-
Christoph Hellwig authored
Signed-off-by:
-
Christoph Hellwig authored
Split out a xlog_add_buffer_cancelled helper which does the low-level manipulation of the buffer cancelation table, and in that helper call xlog_find_buffer_cancelled instead of open coding it. Signed-off-by:
-
Christoph Hellwig authored
Don't bother to allocate memory and convert the log item when we only need the block number and the length. Just extract them directly and call xlog_buf_readahead separately in each branch. Signed-off-by:
-
Christoph Hellwig authored
Add a little helper to readahead a buffer if it hasn't been cancelled. Signed-off-by:
-
Christoph Hellwig authored
This list contains pretty much everything that is not a buffer. The comment calls it item_list, which is a much better name than inode list, so switch the actual variable name to that as well. Signed-off-by:
-
Christoph Hellwig authored
Replace the somewhat convoluted use of xlog_peek_buffer_cancelled and xlog_check_buffer_cancelled with two obvious helpers: xlog_is_buffer_cancelled, which returns true if there is a buffer in the cancellation table, and xlog_put_buffer_cancelled, which also decrements the reference count of the buffer cancellation table. Both share a little helper to look up the entry. Signed-off-by:
-
Eric Sandeen authored
There are a couple places where we directly call printk_once() and one of them doesn't follow the standard xfs subsystem printk format as a result. #define printk_once variants to go with our existing printk_ratelimited #defines so we can do one-shot printks in a consistent manner. Signed-off-by:
Eric Sandeen <sandeen@redhat.com> Reviewed-by:
-
Arnd Bergmann authored
I ran into a linker warning in XFS that originates from a mismatch between libelf, binutils and objtool when certain files in the kernel are built with "gcc -g": x86_64-linux-ld: fs/xfs/xfs_trace.o: unable to initialize decompress status for section .debug_info After some discussion, nobody could identify why xfs sets this flag here. CONFIG_XFS_DEBUG used to enable lots of unrelated settings, but now its main purpose is to enable extra consistency checks and assertions that are unrelated to the debug info. Remove the Makefile logic to set the flag here. If anyone relies on the debug info, this can simply be enabled again with the global CONFIG_DEBUG_INFO option. Dave Chinner writes: I'm pretty sure it was needed for the original kgdb integration back in the early 2000s. That was when SGI used to patch their XFS dev tree with kgdb and debug symbols were needed by the custom kgdb modules that were ported across from the Irix kernel debugger. ISTR that the early kcrash kernel dump analysis tools (again, originated from the Irix "icrash" kernel dump tools) had custom XFS debug scripts that needed also the debug info to work correctly... Which is a long way of saying "we don't need it anymore" instead of "nobody knows why it was set"... :) Suggested-by:
Christoph Hellwig <hch@infradead.org> Link: https://lore.kernel.org/lkml/20200409074130.GD21033@infradead.org/Signed-off-by:
Arnd Bergmann <arnd@arndb.de> Reviewed-by:
-
Kaixu Xia authored
Since the "no-allocation" reservations has been removed, the resblks value should be larger than zero, so remove the unnecessary check. Signed-off-by:
Kaixu Xia <kaixuxia@tencent.com> Reviewed-by:
-
Kaixu Xia authored
Simplify the setting of the flags value, and only consider quota enforcement stuff here. Signed-off-by:
-
Kaixu Xia authored
The check XFS_IS_QUOTA_RUNNING() has been done when enter the xfs_qm_vop_create_dqattach() function, it will return directly if the result is false, so the followed XFS_IS_QUOTA_RUNNING() assertion is unnecessary. If we truly care about this, the check also can be added to the condition of next if statements. Signed-off-by:
-
Kaixu Xia authored
The initial value of variable udqp is NULL, and we only set the flag XFS_QMOPT_PQUOTA in xfs_qm_vop_dqalloc() function, so only the pdqp value is initialized and the udqp value is still NULL. Since the udqp value is NULL in the rest part of xfs_ioctl_setattr() function, it is meaningless and do nothing. So remove it from xfs_ioctl_setattr(). Signed-off-by:
-
