- 29 Jun, 2019 40 commits
-
-
Brian Foster authored
xfs_alloc_ag_vextent_small() doesn't update the output parameters in the event of an AGFL allocation. Instead, it updates the xfs_alloc_arg structure directly to complete the allocation. Update both args and the output params to provide consistent behavior for future callers. Signed-off-by:
Brian Foster <bfoster@redhat.com> Reviewed-by:
Christoph Hellwig <hch@lst.de> Reviewed-by:
Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by:
-
Brian Foster authored
The small allocation helper is implemented in a way that is fairly tightly integrated to the existing allocation algorithms. It expects a cntbt cursor beyond the end of the tree, attempts to locate the last record in the tree and only attempts an AGFL allocation if the cntbt is empty. The upcoming generic algorithm doesn't rely on the cntbt processing of this function. It will only call this function when the cntbt doesn't have a big enough extent or is empty and thus AGFL allocation is the only remaining option. Tweak xfs_alloc_ag_vextent_small() to handle a NULL cntbt cursor and skip the cntbt logic. This facilitates use by the existing allocation code and new code that only requires an AGFL allocation attempt. Signed-off-by:
-
Brian Foster authored
Move the small allocation helper further up in the file to avoid the need for a function declaration. The remaining declarations will be removed by followup patches. No functional changes. Signed-off-by:
-
Brian Foster authored
xfs_alloc_ag_vextent_small() is kind of a mess. Clean it up in preparation for future changes. No functional changes. Signed-off-by:
-
Christoph Hellwig authored
Keep all bmap item related code together. Signed-off-by:
-
Christoph Hellwig authored
Keep all rmap item related code together in one file. Signed-off-by:
-
Christoph Hellwig authored
Keep all the refcount item related code together in one file. Signed-off-by:
-
Christoph Hellwig authored
Keep all the extree item related code together in one file. Signed-off-by:
-
Christoph Hellwig authored
There is no good reason to keep these two functions separate. Signed-off-by:
-
Christoph Hellwig authored
There is no good reason to keep these two functions separate. Signed-off-by:
-
Christoph Hellwig authored
There is no good reason to keep these two functions separate. Signed-off-by:
-
Christoph Hellwig authored
There is no good reason to keep these two functions separate. Signed-off-by:
-
Christoph Hellwig authored
Signed-off-by:
-
Christoph Hellwig authored
Replace the hand grown linked list handling and cil context attachment with the standard list_head structure. Signed-off-by:
-
Christoph Hellwig authored
Signed-off-by:
-
Christoph Hellwig authored
The cast is not type safe, and we can just dereference the first member instead to start with. Signed-off-by:
-
Christoph Hellwig authored
We have various items that are released from ->iop_comitting. Add a flag to just call ->iop_release from the commit path to avoid tons of boilerplate code. Signed-off-by:
-
Christoph Hellwig authored
The iop_unlock method is called when comitting or cancelling a transaction. In the latter case, the transaction may or may not be aborted. While there is no known problem with the current code in practice, this implementation is limited in that any log item implementation that might want to differentiate between a commit and a cancellation must rely on the aborted state. The aborted bit is only set when the cancelled transaction is dirty, however. This means that there is no way to distinguish between a commit and a clean transaction cancellation. For example, intent log items currently rely on this distinction. The log item is either transferred to the CIL on commit or released on transaction cancel. There is currently no possibility for a clean intent log item in a transaction, but if that state is ever introduced a cancel of such a transaction will immediately result in memory leaks of the associated log item(s). This is an interface deficiency and landmine. To clean this up, replace the iop_unlock method with an iop_release method that is specific to transaction cancel. The existing iop_committing method occurs at the same time as iop_unlock in the commit path and there is no need for two separate callbacks here. Overload the iop_committing method with the current commit time iop_unlock implementations to eliminate the need for the latter and further simplify the interface. Signed-off-by:
-
Christoph Hellwig authored
While commiting items looks very similar to freeing them on error it is a different operation, and they will diverge a bit soon. Split out the commit case from xfs_trans_free_items, inline it into xfs_log_commit_cil and give it a separate trace point. Signed-off-by:
-
Christoph Hellwig authored
This method should never be called, so don't waste code on it. Signed-off-by:
-
Christoph Hellwig authored
Just check if they are present first. Signed-off-by:
-
Christoph Hellwig authored
Just pass a straight bool aborted instead of abusing XFS_LI_ABORTED as a flag in function parameters. Signed-off-by:
-
Christoph Hellwig authored
Signed-off-by:
-
Christoph Hellwig authored
We need to derive the mount pointer from a buffer in a lot of place. Add a direct pointer to short cut the pointer chasing. Signed-off-by:
-
Christoph Hellwig authored
This field is now always idential to b_length. Signed-off-by:
Dave Chinner <dchinner@redhat.com> Reviewed-by:
-
Christoph Hellwig authored
Now that the log code doesn't abuse this field any more we can declare it as a struct xfs_buf_log_item pointer. Signed-off-by:
-
Christoph Hellwig authored
Now that the log code uses bios directly we can drop various special cases in the buffer cache code. Signed-off-by:
-
Christoph Hellwig authored
Now that we don't use struct xfs_buf to hold log recovery buffer rename the related functions and variables to just talk of a buffer instead of using the bp name that we usually use for xfs_buf related functionality. Signed-off-by:
-
Christoph Hellwig authored
The xfs_buf structure is basically used as a glorified container for a memory allocation in the log recovery code. Replace it with a call to kmem_alloc_large and a simple abstraction to read into or write from it synchronously using chained bios. Signed-off-by:
-
Christoph Hellwig authored
This simplifies both the helper and the callers. We lost a bit of size sanity checking, but that is already covered by KASAN if needed. Signed-off-by:
-
Christoph Hellwig authored
Move the workqueue used for log I/O completions from struct xfs_mount to struct xlog to keep it self contained in the log code. Signed-off-by:
-
Christoph Hellwig authored
Currently the XFS logging code uses the xfs_buf structure and associated APIs to write the log buffers to disk. This requires various special cases in the log code and is generally not very optimal. Instead of using a buffer just allocate a kmem_alloc_larger region for each log buffer, and use a bio and bio_vec array embedded in the iclog structure to write the buffer to disk. This also allows for using the bio split and chaining case to deal with the case of a log buffer wrapping around the end of the log. Signed-off-by:
-
Christoph Hellwig authored
Use the slightly shorter way to get at the buftarg for the log device wherever we can in the log and log recovery code. Signed-off-by:
-
Christoph Hellwig authored
The only caller unconditionally passes true here. Signed-off-by:
-
Christoph Hellwig authored
Just a small bit of code tidying up. Signed-off-by:
-
Christoph Hellwig authored
Split out another self-contained bit of code from xlog_sync. Signed-off-by:
-
Christoph Hellwig authored
Split out a self-contained chunk of code from xlog_sync that calculates the split offset for an iclog that wraps the log end and bumps the cycles for the second half. Use the chance to bring some sanity to the variables used to track the split in xlog_sync by not changing the count variable, and instead use split as the offset for the split and use those to calculate the sizes and offsets for the two write buffers. Signed-off-by:
-
Christoph Hellwig authored
Replace the not very useful xlog_bdstrat wrapper with a new version that that takes care of all the common logic for writing log buffers. Use the opportunity to avoid overloading the buffer address with the log relative address, and to shed the unused return value. Signed-off-by:
-
Christoph Hellwig authored
If we have to split a log write because it wraps the end of the log we can't just use REQ_PREFLUSH to flush before the first log write, as the writes might get reordered somewhere in the I/O stack. Issue a manual flush in that case so that the ordering of the two log I/Os doesn't matter. Signed-off-by:
-
Christoph Hellwig authored
This value is the only flag in ic_state, which we otherwise use as a state. Switch it to a new debug-only field and also report and actual error in the buffer in the I/O completion path. Signed-off-by:
-
