| Message ID | a45bd8eb8d16b648368b2e83f12a72ea44dab71c.1713937746.git.wqu@suse.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | btrfs-progs: fsfeatures: hide RST behind experimental builds | expand |
Looks good,
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Although I thought that you need to configure btrfs-progs with
--enable-experimental to get RST support :/
On 24/04/2024 07.49, Qu Wenruo wrote: > [BUG] > There is a report that with v6.7.1 btrfs-progs, `mkfs.btrfs -O rst`, but > kernel 6.7/6.8 are unable to mount it at all. > > [CAUSE] > Although the feature string states the raid-stripe-tree feature is > supported since v6.7 kernel, it's not correct. > Only debug kernel with CONFIG_BTRFS_DEBUG would support the RST feature. > > Thus RST is still considered experimental. > > [FIX] > Move the RST mkfs features back to experimental. > > This patch would only hide the RST features from 'mkfs -O' options, the > existing supporting code for RST would still be there, thus > non-experimental build of `btrfs check` can still verify a btrfs with > RST. > > Reported-by: HAN Yuwei <hrx@bupt.moe> > Fixes: b421fdff9574 ("btrfs-progs: move raid-stripe-tree and squota build out of experimental") > Signed-off-by: Qu Wenruo <wqu@suse.com> > --- > common/fsfeatures.c | 2 ++ > 1 file changed, 2 insertions(+) > > diff --git a/common/fsfeatures.c b/common/fsfeatures.c > index c5e81629ccea..7aaddab6a192 100644 > --- a/common/fsfeatures.c > +++ b/common/fsfeatures.c > @@ -222,6 +222,7 @@ static const struct btrfs_feature mkfs_features[] = { > VERSION_NULL(default), > .desc = "block group tree, more efficient block group tracking to reduce mount time" > }, > +#if EXPERIMENTAL > { > .name = "rst", > .incompat_flag = BTRFS_FEATURE_INCOMPAT_RAID_STRIPE_TREE, > @@ -238,6 +239,7 @@ static const struct btrfs_feature mkfs_features[] = { > VERSION_NULL(default), > .desc = "raid stripe tree, enhanced file extent tracking" > }, > +#endif > { > .name = "squota", > .incompat_flag = BTRFS_FEATURE_INCOMPAT_SIMPLE_QUOTA, I am bit confused. The Han report say that the problem is due to the fact that the option is enabled by *default*. So way we should remove (from the binary) this option at all ? Shouldn't be enough to remove it from the options enabled by default ? Something like: - VERSION_TO_STRING2(compat, 6,7), - VERSION_NULL(safe), + VERSION_NULL(compat), + VERSION_TO_STRING2(safe, 6,7),
On 24/04/2024 21.30, Goffredo Baroncelli wrote: > On 24/04/2024 07.49, Qu Wenruo wrote: >> [BUG] >> There is a report that with v6.7.1 btrfs-progs, `mkfs.btrfs -O rst`, but >> kernel 6.7/6.8 are unable to mount it at all. >> >> [CAUSE] >> Although the feature string states the raid-stripe-tree feature is >> supported since v6.7 kernel, it's not correct. >> Only debug kernel with CONFIG_BTRFS_DEBUG would support the RST feature. >> >> Thus RST is still considered experimental. >> >> [FIX] >> Move the RST mkfs features back to experimental. >> >> This patch would only hide the RST features from 'mkfs -O' options, the >> existing supporting code for RST would still be there, thus >> non-experimental build of `btrfs check` can still verify a btrfs with >> RST. >> >> Reported-by: HAN Yuwei <hrx@bupt.moe> >> Fixes: b421fdff9574 ("btrfs-progs: move raid-stripe-tree and squota build out of experimental") >> Signed-off-by: Qu Wenruo <wqu@suse.com> >> --- >> common/fsfeatures.c | 2 ++ >> 1 file changed, 2 insertions(+) >> >> diff --git a/common/fsfeatures.c b/common/fsfeatures.c >> index c5e81629ccea..7aaddab6a192 100644 >> --- a/common/fsfeatures.c >> +++ b/common/fsfeatures.c >> @@ -222,6 +222,7 @@ static const struct btrfs_feature mkfs_features[] = { >> VERSION_NULL(default), >> .desc = "block group tree, more efficient block group tracking to reduce mount time" >> }, >> +#if EXPERIMENTAL >> { >> .name = "rst", >> .incompat_flag = BTRFS_FEATURE_INCOMPAT_RAID_STRIPE_TREE, >> @@ -238,6 +239,7 @@ static const struct btrfs_feature mkfs_features[] = { >> VERSION_NULL(default), >> .desc = "raid stripe tree, enhanced file extent tracking" >> }, >> +#endif >> { >> .name = "squota", >> .incompat_flag = BTRFS_FEATURE_INCOMPAT_SIMPLE_QUOTA, > > I am bit confused. Ignore this email. > > The Han report say that the problem is due to the fact that the option is enabled by *default*. > > So way we should remove (from the binary) this option at all ? Shouldn't be enough to remove it from the options enabled by default ? > > Something like: > > - VERSION_TO_STRING2(compat, 6,7), > - VERSION_NULL(safe), > > + VERSION_NULL(compat), > > + VERSION_TO_STRING2(safe, 6,7), > >
diff --git a/common/fsfeatures.c b/common/fsfeatures.c index c5e81629ccea..7aaddab6a192 100644 --- a/common/fsfeatures.c +++ b/common/fsfeatures.c @@ -222,6 +222,7 @@ static const struct btrfs_feature mkfs_features[] = { VERSION_NULL(default), .desc = "block group tree, more efficient block group tracking to reduce mount time" }, +#if EXPERIMENTAL { .name = "rst", .incompat_flag = BTRFS_FEATURE_INCOMPAT_RAID_STRIPE_TREE, @@ -238,6 +239,7 @@ static const struct btrfs_feature mkfs_features[] = { VERSION_NULL(default), .desc = "raid stripe tree, enhanced file extent tracking" }, +#endif { .name = "squota", .incompat_flag = BTRFS_FEATURE_INCOMPAT_SIMPLE_QUOTA,
[BUG] There is a report that with v6.7.1 btrfs-progs, `mkfs.btrfs -O rst`, but kernel 6.7/6.8 are unable to mount it at all. [CAUSE] Although the feature string states the raid-stripe-tree feature is supported since v6.7 kernel, it's not correct. Only debug kernel with CONFIG_BTRFS_DEBUG would support the RST feature. Thus RST is still considered experimental. [FIX] Move the RST mkfs features back to experimental. This patch would only hide the RST features from 'mkfs -O' options, the existing supporting code for RST would still be there, thus non-experimental build of `btrfs check` can still verify a btrfs with RST. Reported-by: HAN Yuwei <hrx@bupt.moe> Fixes: b421fdff9574 ("btrfs-progs: move raid-stripe-tree and squota build out of experimental") Signed-off-by: Qu Wenruo <wqu@suse.com> --- common/fsfeatures.c | 2 ++ 1 file changed, 2 insertions(+)