Message ID | 20240229113849.2222577-2-chengzhihao1@huawei.com (mailing list archive) |
---|---|
State | New |
Headers | show |
Series | ext4: Do endio process under irq context for DIO overwrites | expand |
On Thu, Feb 29, 2024 at 07:38:48PM +0800, Zhihao Cheng wrote: > It will be more efficient to execute quick endio process(eg. non-sync > overwriting case) under irq process rather than starting a worker to > do it. > Add a flag to control DIO to be finished inline(under irq context), which > can be used for non-sync overwriting case. > Besides, skip invalidating pages if DIO is finished inline, which will > keep the same logic with dio_bio_end_aio in non-sync overwriting case. > > Signed-off-by: Zhihao Cheng <chengzhihao1@huawei.com> A nice idea, but I don't think an ext4 specific API flag is the right way to go about enabling this. The iomap dio code knows if the write is pure overwrite already - we have the IOMAP_F_DIRTY flag for that, and we combine this with IOMAP_DIO_WRITE_THROUGH to do the pure overwrite FUA optimisations. That is: /* * Use a FUA write if we need datasync semantics, this is a pure * data IO that doesn't require any metadata updates (including * after IO completion such as unwritten extent conversion) and * the underlying device either supports FUA or doesn't have * a volatile write cache. This allows us to avoid cache flushes * on IO completion. If we can't use writethrough and need to * sync, disable in-task completions as dio completion will * need to call generic_write_sync() which will do a blocking * fsync / cache flush call. */ if (!(iomap->flags & (IOMAP_F_SHARED|IOMAP_F_DIRTY)) && (dio->flags & IOMAP_DIO_WRITE_THROUGH) && (bdev_fua(iomap->bdev) || !bdev_write_cache(iomap->bdev))) use_fua = true; Hence if we want to optimise pure overwrites that have no data sync requirements, we already have the detection and triggers in place to do this. We just need to change the way we set up the IO flags to allow write-through (i.e. non-blocking IO completions) to use inline completions. In __iomap_dio_rw(): + /* Always try to complete inline. */ + dio->flags |= IOMAP_DIO_INLINE_COMP; if (iov_iter_rw(iter) == READ) { - /* reads can always complete inline */ - dio->flags |= IOMAP_DIO_INLINE_COMP; .... } else { + /* Always try write-through semantics. If we can't + * use writethough, it will be disabled along with + * IOMAP_DIO_INLINE_COMP before dio completion is run + * so it can be deferred to a task completion context + * appropriately. + */ + dio->flags |= IOMAP_DIO_WRITE | IOMAP_DIO_WRITE_THROUGH; iomi.flags |= IOMAP_WRITE; - dio->flags |= IOMAP_DIO_WRITE; ..... /* for data sync or sync, we need sync completion processing */ if (iocb_is_dsync(iocb)) { dio->flags |= IOMAP_DIO_NEED_SYNC; - /* - * For datasync only writes, we optimistically try using - * WRITE_THROUGH for this IO. This flag requires either - * FUA writes through the device's write cache, or a - * normal write to a device without a volatile write - * cache. For the former, Any non-FUA write that occurs - * will clear this flag, hence we know before completion - * whether a cache flush is necessary. - */ - if (!(iocb->ki_flags & IOCB_SYNC)) - dio->flags |= IOMAP_DIO_WRITE_THROUGH; + * For sync writes we know we are going to need + * blocking completion processing, so turn off + * writethrough now. + */ if (iocb->ki_flags & IOCB_SYNC) { dio->flags &= ~(IOMAP_DIO_WRITE_THROUGH | IOMAP_DIO_INLINE_COMP); } } This then sets up iomap_dio_bio_iter() to be able to determine if the iomap returned is for a pure overwrite and allow for the use of inline write completions. /* * If we have a pure overwrite, we know that IO completion * will not block and so we can use write through completion * semantics and complete the write inline. If it's not a * pure overwrite, make sure that we always defer * completions to a task context. */ if (dio->flags & IOMAP_DIO_WRITE_THROUGH) { if (iomap->flags & (IOMAP_F_SHARED|IOMAP_F_DIRTY)) { dio->flags &= ~(IOMAP_DIO_WRITE_THROUGH | IOMAP_DIO_INLINE_COMP); } else if ((dio->flags & IOMAP_DIO_NEED_SYNC) && (bdev_fua(iomap->bdev) || !bdev_write_cache(iomap->bdev))) { /* * Use REQ_FUA for datasync overwrites to * avoid cache flushes on IO completion on * devices that support FUA or don't have * volatile caches. */ use_fua = true; } } and iomap_dio_bio_opflags() gets changed to also clear IOMAP_DIO_INLINE_COMP when it clears IOMAP_DIO_WRITE_THROUGH.... That then makes all pure overwrites for every filesystem do inline completions without changing calling conventions. i.e. it's up to the filesystem ->iomap_begin callouts to indicate whether the write mapping returned requires blocking operations to be done in IO completion (i.e. set the IOMAP_F_DIRTY flag) appropriately. However, this does mean that any spinlock taken in the ->end_io() callbacks now needs to be irq safe. e.g. in xfs_dio_write_end_io() the spinlock protection around inode size updates will need to use an irq safe locking, as will the locking in the DIO submission path that it serialises against in xfs_file_write_checks(). That probably is best implemented as a separate spinlock. There will also be other filesystems that need to set IOMAP_F_DIRTY unconditionally (e.g. zonefs) because they always take blocking locks in their ->end_io callbacks and so must always run in task context... Cheers, Dave.
在 2024/3/1 8:40, Dave Chinner 写道: Hi Dave, thanks for your detailed and nice suggestions, I have a few questions below. > On Thu, Feb 29, 2024 at 07:38:48PM +0800, Zhihao Cheng wrote: >> It will be more efficient to execute quick endio process(eg. non-sync >> overwriting case) under irq process rather than starting a worker to >> do it. >> Add a flag to control DIO to be finished inline(under irq context), which >> can be used for non-sync overwriting case. >> Besides, skip invalidating pages if DIO is finished inline, which will >> keep the same logic with dio_bio_end_aio in non-sync overwriting case. >> >> Signed-off-by: Zhihao Cheng <chengzhihao1@huawei.com> > > A nice idea, but I don't think an ext4 specific API flag is the > right way to go about enabling this. The iomap dio code knows if > the write is pure overwrite already - we have the IOMAP_F_DIRTY flag > for that, and we combine this with IOMAP_DIO_WRITE_THROUGH to do the > pure overwrite FUA optimisations. > > That is: > > /* > * Use a FUA write if we need datasync semantics, this is a pure > * data IO that doesn't require any metadata updates (including > * after IO completion such as unwritten extent conversion) and > * the underlying device either supports FUA or doesn't have > * a volatile write cache. This allows us to avoid cache flushes > * on IO completion. If we can't use writethrough and need to > * sync, disable in-task completions as dio completion will > * need to call generic_write_sync() which will do a blocking > * fsync / cache flush call. > */ > if (!(iomap->flags & (IOMAP_F_SHARED|IOMAP_F_DIRTY)) && > (dio->flags & IOMAP_DIO_WRITE_THROUGH) && > (bdev_fua(iomap->bdev) || !bdev_write_cache(iomap->bdev))) > use_fua = true; > > Hence if we want to optimise pure overwrites that have no data sync > requirements, we already have the detection and triggers in place to > do this. We just need to change the way we set up the IO flags to > allow write-through (i.e. non-blocking IO completions) to use inline > completions. > > In __iomap_dio_rw(): > > + /* Always try to complete inline. */ > + dio->flags |= IOMAP_DIO_INLINE_COMP; > if (iov_iter_rw(iter) == READ) { > - /* reads can always complete inline */ > - dio->flags |= IOMAP_DIO_INLINE_COMP; > .... > > } else { > + /* Always try write-through semantics. If we can't > + * use writethough, it will be disabled along with > + * IOMAP_DIO_INLINE_COMP before dio completion is run > + * so it can be deferred to a task completion context > + * appropriately. > + */ > + dio->flags |= IOMAP_DIO_WRITE | IOMAP_DIO_WRITE_THROUGH; There is a behavior change here, if we set IOMAP_DIO_WRITE_THROUGH unconditionally, non-datasync IO will be set with REQ_FUA, which means that device will flush writecache for each IO, will it affect the performance in non-sync dio case? > iomi.flags |= IOMAP_WRITE; > - dio->flags |= IOMAP_DIO_WRITE; > ..... > /* for data sync or sync, we need sync completion processing */ > if (iocb_is_dsync(iocb)) { > dio->flags |= IOMAP_DIO_NEED_SYNC; > > - /* > - * For datasync only writes, we optimistically try using > - * WRITE_THROUGH for this IO. This flag requires either > - * FUA writes through the device's write cache, or a > - * normal write to a device without a volatile write > - * cache. For the former, Any non-FUA write that occurs > - * will clear this flag, hence we know before completion > - * whether a cache flush is necessary. > - */ > - if (!(iocb->ki_flags & IOCB_SYNC)) > - dio->flags |= IOMAP_DIO_WRITE_THROUGH; > + * For sync writes we know we are going to need > + * blocking completion processing, so turn off > + * writethrough now. > + */ > if (iocb->ki_flags & IOCB_SYNC) { > dio->flags &= ~(IOMAP_DIO_WRITE_THROUGH | > IOMAP_DIO_INLINE_COMP); > } > } > [...] > > However, this does mean that any spinlock taken in the ->end_io() > callbacks now needs to be irq safe. e.g. in xfs_dio_write_end_io() > the spinlock protection around inode size updates will need to use > an irq safe locking, as will the locking in the DIO submission path > that it serialises against in xfs_file_write_checks(). That probably > is best implemented as a separate spinlock. > > There will also be other filesystems that need to set IOMAP_F_DIRTY > unconditionally (e.g. zonefs) because they always take blocking > locks in their ->end_io callbacks and so must always run in task > context... Should we add a new flag(eg. IOMAP_F_ENDIO_IRQ ?) to indicate that the endio cannot be done under irq? Because I think IOMAP_F_DIRTY means that the metadata needs to be written, if we add a new semantics(endio must be done in defered work) for this flag, the code will looks a little complicated.
在 2024/3/1 18:02, Zhihao Cheng 写道: > 在 2024/3/1 8:40, Dave Chinner 写道: > Hi Dave. Friendly ping. > Hi Dave, thanks for your detailed and nice suggestions, I have a few > questions below. >> On Thu, Feb 29, 2024 at 07:38:48PM +0800, Zhihao Cheng wrote: >>> It will be more efficient to execute quick endio process(eg. non-sync >>> overwriting case) under irq process rather than starting a worker to >>> do it. >>> Add a flag to control DIO to be finished inline(under irq context), >>> which >>> can be used for non-sync overwriting case. >>> Besides, skip invalidating pages if DIO is finished inline, which will >>> keep the same logic with dio_bio_end_aio in non-sync overwriting case. >>> >>> Signed-off-by: Zhihao Cheng <chengzhihao1@huawei.com> >> >> A nice idea, but I don't think an ext4 specific API flag is the >> right way to go about enabling this. The iomap dio code knows if >> the write is pure overwrite already - we have the IOMAP_F_DIRTY flag >> for that, and we combine this with IOMAP_DIO_WRITE_THROUGH to do the >> pure overwrite FUA optimisations. >> >> That is: >> >> /* >> * Use a FUA write if we need datasync semantics, >> this is a pure >> * data IO that doesn't require any metadata updates >> (including >> * after IO completion such as unwritten extent >> conversion) and >> * the underlying device either supports FUA or >> doesn't have >> * a volatile write cache. This allows us to avoid >> cache flushes >> * on IO completion. If we can't use writethrough and >> need to >> * sync, disable in-task completions as dio >> completion will >> * need to call generic_write_sync() which will do a >> blocking >> * fsync / cache flush call. >> */ >> if (!(iomap->flags & (IOMAP_F_SHARED|IOMAP_F_DIRTY)) && >> (dio->flags & IOMAP_DIO_WRITE_THROUGH) && >> (bdev_fua(iomap->bdev) || >> !bdev_write_cache(iomap->bdev))) >> use_fua = true; >> >> Hence if we want to optimise pure overwrites that have no data sync >> requirements, we already have the detection and triggers in place to >> do this. We just need to change the way we set up the IO flags to >> allow write-through (i.e. non-blocking IO completions) to use inline >> completions. >> >> In __iomap_dio_rw(): >> >> + /* Always try to complete inline. */ >> + dio->flags |= IOMAP_DIO_INLINE_COMP; >> if (iov_iter_rw(iter) == READ) { >> - /* reads can always complete inline */ >> - dio->flags |= IOMAP_DIO_INLINE_COMP; >> .... >> >> } else { >> + /* Always try write-through semantics. If we can't >> + * use writethough, it will be disabled along with >> + * IOMAP_DIO_INLINE_COMP before dio completion is run >> + * so it can be deferred to a task completion context >> + * appropriately. >> + */ >> + dio->flags |= IOMAP_DIO_WRITE | IOMAP_DIO_WRITE_THROUGH; > > There is a behavior change here, if we set IOMAP_DIO_WRITE_THROUGH > unconditionally, non-datasync IO will be set with REQ_FUA, which means > that device will flush writecache for each IO, will it affect the > performance in non-sync dio case? >> iomi.flags |= IOMAP_WRITE; >> - dio->flags |= IOMAP_DIO_WRITE; >> ..... >> /* for data sync or sync, we need sync completion processing */ >> if (iocb_is_dsync(iocb)) { >> dio->flags |= IOMAP_DIO_NEED_SYNC; >> >> - /* >> - * For datasync only writes, we optimistically >> try using >> - * WRITE_THROUGH for this IO. This flag >> requires either >> - * FUA writes through the device's write cache, >> or a >> - * normal write to a device without a volatile >> write >> - * cache. For the former, Any non-FUA write >> that occurs >> - * will clear this flag, hence we know before >> completion >> - * whether a cache flush is necessary. >> - */ >> - if (!(iocb->ki_flags & IOCB_SYNC)) >> - dio->flags |= IOMAP_DIO_WRITE_THROUGH; >> + * For sync writes we know we are going to need >> + * blocking completion processing, so turn off >> + * writethrough now. >> + */ >> if (iocb->ki_flags & IOCB_SYNC) { >> dio->flags &= ~(IOMAP_DIO_WRITE_THROUGH | >> IOMAP_DIO_INLINE_COMP); >> } >> } >> > > [...] >> >> However, this does mean that any spinlock taken in the ->end_io() >> callbacks now needs to be irq safe. e.g. in xfs_dio_write_end_io() >> the spinlock protection around inode size updates will need to use >> an irq safe locking, as will the locking in the DIO submission path >> that it serialises against in xfs_file_write_checks(). That probably >> is best implemented as a separate spinlock. >> >> There will also be other filesystems that need to set IOMAP_F_DIRTY >> unconditionally (e.g. zonefs) because they always take blocking >> locks in their ->end_io callbacks and so must always run in task >> context... > Should we add a new flag(eg. IOMAP_F_ENDIO_IRQ ?) to indicate that the > endio cannot be done under irq? Because I think IOMAP_F_DIRTY means that > the metadata needs to be written, if we add a new semantics(endio must > be done in defered work) for this flag, the code will looks a little > complicated. > > > .
diff --git a/fs/iomap/direct-io.c b/fs/iomap/direct-io.c index bcd3f8cf5ea4..221715b38ce2 100644 --- a/fs/iomap/direct-io.c +++ b/fs/iomap/direct-io.c @@ -110,7 +110,8 @@ ssize_t iomap_dio_complete(struct iomap_dio *dio) * ->end_io() when necessary, otherwise a racing buffer read would cache * zeros from unwritten extents. */ - if (!dio->error && dio->size && (dio->flags & IOMAP_DIO_WRITE)) + if (!dio->error && dio->size && (dio->flags & IOMAP_DIO_WRITE) && + !(dio->flags & IOMAP_DIO_INLINE_COMP)) kiocb_invalidate_post_direct_write(iocb, dio->size); inode_dio_end(file_inode(iocb->ki_filp)); @@ -122,8 +123,10 @@ ssize_t iomap_dio_complete(struct iomap_dio *dio) * If this is a DSYNC write, make sure we push it to stable * storage now that we've written data. */ - if (dio->flags & IOMAP_DIO_NEED_SYNC) + if (dio->flags & IOMAP_DIO_NEED_SYNC) { + WARN_ON_ONCE(dio->flags & IOMAP_DIO_INLINE_COMP); ret = generic_write_sync(iocb, ret); + } if (ret > 0) ret += dio->done_before; } @@ -628,6 +631,9 @@ __iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter, */ if (!(iocb->ki_flags & IOCB_SYNC)) dio->flags |= IOMAP_DIO_WRITE_THROUGH; + } else if (dio_flags & IOMAP_DIO_MAY_INLINE_COMP) { + /* writes could complete inline */ + dio->flags |= IOMAP_DIO_INLINE_COMP; } /* diff --git a/include/linux/iomap.h b/include/linux/iomap.h index 96dd0acbba44..f292b10028d0 100644 --- a/include/linux/iomap.h +++ b/include/linux/iomap.h @@ -382,6 +382,12 @@ struct iomap_dio_ops { */ #define IOMAP_DIO_PARTIAL (1 << 2) +/* + * DIO will be completed inline unless sync operation is needed after io is + * finished. + */ +#define IOMAP_DIO_MAY_INLINE_COMP (1 << 3) + ssize_t iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter, const struct iomap_ops *ops, const struct iomap_dio_ops *dops, unsigned int dio_flags, void *private, size_t done_before);
It will be more efficient to execute quick endio process(eg. non-sync overwriting case) under irq process rather than starting a worker to do it. Add a flag to control DIO to be finished inline(under irq context), which can be used for non-sync overwriting case. Besides, skip invalidating pages if DIO is finished inline, which will keep the same logic with dio_bio_end_aio in non-sync overwriting case. Signed-off-by: Zhihao Cheng <chengzhihao1@huawei.com> --- fs/iomap/direct-io.c | 10 ++++++++-- include/linux/iomap.h | 6 ++++++ 2 files changed, 14 insertions(+), 2 deletions(-)