| Message ID | 20240213093713.1753368-14-kernel@pankajraghav.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | enable bs > ps in XFS | expand |
On Tue, Feb 13, 2024 at 10:37:12AM +0100, Pankaj Raghav (Samsung) wrote: > From: Pankaj Raghav <p.raghav@samsung.com> > > Add an experimental CONFIG_XFS_LBS option to enable LBS support in XFS. > Retain the ASSERT for PAGE_SHIFT if CONFIG_XFS_LBS is not enabled. > > Signed-off-by: Pankaj Raghav <p.raghav@samsung.com> With the changes I suggested in the next patch, Reviewed-by: Darrick J. Wong <djwong@kernel.org> (I think you ought to combine this patch and the next patch after improving the EXPERIMENTAL log messaging.) --D > --- > fs/xfs/Kconfig | 11 +++++++++++ > fs/xfs/xfs_mount.c | 4 +++- > 2 files changed, 14 insertions(+), 1 deletion(-) > > diff --git a/fs/xfs/Kconfig b/fs/xfs/Kconfig > index 567fb37274d3..6b0db2f7dc13 100644 > --- a/fs/xfs/Kconfig > +++ b/fs/xfs/Kconfig > @@ -216,3 +216,14 @@ config XFS_ASSERT_FATAL > result in warnings. > > This behavior can be modified at runtime via sysfs. > + > +config XFS_LBS > + bool "XFS large block size support (EXPERIMENTAL)" > + depends on XFS_FS > + help > + Set Y to enable support for filesystem block size > system's > + base page size. > + > + This feature is considered EXPERIMENTAL. Use with caution! > + > + If unsure, say N. > diff --git a/fs/xfs/xfs_mount.c b/fs/xfs/xfs_mount.c > index bfbaaecaf668..596aa2cdefbc 100644 > --- a/fs/xfs/xfs_mount.c > +++ b/fs/xfs/xfs_mount.c > @@ -131,11 +131,13 @@ xfs_sb_validate_fsb_count( > xfs_sb_t *sbp, > uint64_t nblocks) > { > - ASSERT(PAGE_SHIFT >= sbp->sb_blocklog); > ASSERT(sbp->sb_blocklog >= BBSHIFT); > unsigned long mapping_count; > uint64_t bytes = nblocks << sbp->sb_blocklog; > > + if (!IS_ENABLED(CONFIG_XFS_LBS)) > + ASSERT(PAGE_SHIFT >= sbp->sb_blocklog); > + > mapping_count = bytes >> PAGE_SHIFT; > > /* Limited by ULONG_MAX of page cache index */ > -- > 2.43.0 > >
On Tue, Feb 13, 2024 at 10:37:12AM +0100, Pankaj Raghav (Samsung) wrote: > From: Pankaj Raghav <p.raghav@samsung.com> > > Add an experimental CONFIG_XFS_LBS option to enable LBS support in XFS. > Retain the ASSERT for PAGE_SHIFT if CONFIG_XFS_LBS is not enabled. > > Signed-off-by: Pankaj Raghav <p.raghav@samsung.com> NAK. There it no reason for this existing - the same code is run regardless of the state of this config variable just with a difference in min folio order. All it does is increase the test matrix arbitrarily - now we have two kernel configs we have to test and there's no good reason for doing that. -Dave.
On Wed, Feb 14, 2024 at 08:19:23AM +1100, Dave Chinner wrote: > On Tue, Feb 13, 2024 at 10:37:12AM +0100, Pankaj Raghav (Samsung) wrote: > > From: Pankaj Raghav <p.raghav@samsung.com> > > > > Add an experimental CONFIG_XFS_LBS option to enable LBS support in XFS. > > Retain the ASSERT for PAGE_SHIFT if CONFIG_XFS_LBS is not enabled. > > > > Signed-off-by: Pankaj Raghav <p.raghav@samsung.com> > > NAK. > > There it no reason for this existing - the same code is run > regardless of the state of this config variable just with a > difference in min folio order. All it does is increase the test > matrix arbitrarily - now we have two kernel configs we have to test > and there's no good reason for doing that. I did not have this CONFIG in the first round but I thought it might help retain the existing behaviour until we deem the feature stable. But I get your point. So we remove this CONFIG and just have an experimental warning during mount when people are using the LBS support? > > -Dave. > -- > Dave Chinner > david@fromorbit.com
On Tue, Feb 13, 2024 at 10:54:07PM +0100, Pankaj Raghav (Samsung) wrote: > On Wed, Feb 14, 2024 at 08:19:23AM +1100, Dave Chinner wrote: > > On Tue, Feb 13, 2024 at 10:37:12AM +0100, Pankaj Raghav (Samsung) wrote: > > > From: Pankaj Raghav <p.raghav@samsung.com> > > > > > > Add an experimental CONFIG_XFS_LBS option to enable LBS support in XFS. > > > Retain the ASSERT for PAGE_SHIFT if CONFIG_XFS_LBS is not enabled. > > > > > > Signed-off-by: Pankaj Raghav <p.raghav@samsung.com> > > > > NAK. > > > > There it no reason for this existing - the same code is run > > regardless of the state of this config variable just with a > > difference in min folio order. All it does is increase the test > > matrix arbitrarily - now we have two kernel configs we have to test > > and there's no good reason for doing that. > > I did not have this CONFIG in the first round but I thought it might > help retain the existing behaviour until we deem the feature stable. > > But I get your point. So we remove this CONFIG and just have an > experimental warning during mount when people are using the LBS support? Yes. -Dave.
diff --git a/fs/xfs/Kconfig b/fs/xfs/Kconfig index 567fb37274d3..6b0db2f7dc13 100644 --- a/fs/xfs/Kconfig +++ b/fs/xfs/Kconfig @@ -216,3 +216,14 @@ config XFS_ASSERT_FATAL result in warnings. This behavior can be modified at runtime via sysfs. + +config XFS_LBS + bool "XFS large block size support (EXPERIMENTAL)" + depends on XFS_FS + help + Set Y to enable support for filesystem block size > system's + base page size. + + This feature is considered EXPERIMENTAL. Use with caution! + + If unsure, say N. diff --git a/fs/xfs/xfs_mount.c b/fs/xfs/xfs_mount.c index bfbaaecaf668..596aa2cdefbc 100644 --- a/fs/xfs/xfs_mount.c +++ b/fs/xfs/xfs_mount.c @@ -131,11 +131,13 @@ xfs_sb_validate_fsb_count( xfs_sb_t *sbp, uint64_t nblocks) { - ASSERT(PAGE_SHIFT >= sbp->sb_blocklog); ASSERT(sbp->sb_blocklog >= BBSHIFT); unsigned long mapping_count; uint64_t bytes = nblocks << sbp->sb_blocklog; + if (!IS_ENABLED(CONFIG_XFS_LBS)) + ASSERT(PAGE_SHIFT >= sbp->sb_blocklog); + mapping_count = bytes >> PAGE_SHIFT; /* Limited by ULONG_MAX of page cache index */