| Message ID | ee78c5dd-5ee4-994c-47e2-209e38a9e986@redhat.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | xfs: mark ARM OABI as incompatible in Kconfig | expand |
On Mon, Feb 24, 2020 at 04:49:12PM -0800, Eric Sandeen wrote: > The old ARM OABI's structure alignment quirks break xfs disk structures, > let's just move on and disallow it rather than playing whack-a-mole > for the infrequent times someone selects this old config, which is > usually during "make randconfig" tests. > > Signed-off-by: Eric Sandeen <sandeen@redhat.com> > --- > > diff --git a/fs/xfs/Kconfig b/fs/xfs/Kconfig > index e685299eb3d2..043624bd4ab2 100644 > --- a/fs/xfs/Kconfig > +++ b/fs/xfs/Kconfig > @@ -2,6 +2,8 @@ > config XFS_FS > tristate "XFS filesystem support" > depends on BLOCK > + # We don't support OABI structure alignment on ARM Should this limitation be documented in the help screen? > + depends on (!ARM || AEABI) > select EXPORTFS > select LIBCRC32C > select FS_IOMAP >
On 2/24/20 4:55 PM, Darrick J. Wong wrote: > On Mon, Feb 24, 2020 at 04:49:12PM -0800, Eric Sandeen wrote: >> The old ARM OABI's structure alignment quirks break xfs disk structures, >> let's just move on and disallow it rather than playing whack-a-mole >> for the infrequent times someone selects this old config, which is >> usually during "make randconfig" tests. >> >> Signed-off-by: Eric Sandeen <sandeen@redhat.com> >> --- >> >> diff --git a/fs/xfs/Kconfig b/fs/xfs/Kconfig >> index e685299eb3d2..043624bd4ab2 100644 >> --- a/fs/xfs/Kconfig >> +++ b/fs/xfs/Kconfig >> @@ -2,6 +2,8 @@ >> config XFS_FS >> tristate "XFS filesystem support" >> depends on BLOCK >> + # We don't support OABI structure alignment on ARM > > Should this limitation be documented in the help screen? Yeah probably. But now looking at aa2dd0ad4d6d xfs: remove __arch_pack hch indicates that some non-arm architectures have similar problems, so is there any point to excluding this one config on this one arch? -Eric
On Mon, Feb 24, 2020 at 04:58:20PM -0800, Eric Sandeen wrote: > Yeah probably. > > But now looking at > > aa2dd0ad4d6d xfs: remove __arch_pack > > hch indicates that some non-arm architectures have similar problems, > so is there any point to excluding this one config on this one arch? Yes, I think we just need to be more careful with our newly added structure layouts and whenever we do a major reorganization of an existing one.
diff --git a/fs/xfs/Kconfig b/fs/xfs/Kconfig index e685299eb3d2..043624bd4ab2 100644 --- a/fs/xfs/Kconfig +++ b/fs/xfs/Kconfig @@ -2,6 +2,8 @@ config XFS_FS tristate "XFS filesystem support" depends on BLOCK + # We don't support OABI structure alignment on ARM + depends on (!ARM || AEABI) select EXPORTFS select LIBCRC32C select FS_IOMAP
The old ARM OABI's structure alignment quirks break xfs disk structures, let's just move on and disallow it rather than playing whack-a-mole for the infrequent times someone selects this old config, which is usually during "make randconfig" tests. Signed-off-by: Eric Sandeen <sandeen@redhat.com> ---