On 2018/3/11 16:12, Amir Goldstein Wrote: > On Thu, Mar 1, 2018 at 3:11 PM, zhangyi (F) <yi.zhang@xxxxxxxxxx> wrote: >> On 2018/3/1 21:03, Amir Goldstein Wrote: >>> On Thu, Mar 1, 2018 at 2:13 PM, zhangyi (F) <yi.zhang@xxxxxxxxxx> wrote: >>>> Hook filesystem check helper to _check_test_fs and _check_scratch_fs >>>> for checking consistency of underlying dirs of overlay filesystem. >>>> These helpers works only if fsck.overlay exists. >>>> >>>> This patch introduce OVERLAY_FSCK_OPTIONS use for check overlayfs like >>>> OVERLAY_MOUNT_OPTIONS, and also introduce a mount point check helper in >>>> common/rc to detect a dir is a mount point or not. >>>> >>>> [ _check_test_fs/_check_scratch_fs part picked from Amir's patch, thanks ] >>>> >>>> Signed-off-by: zhangyi (F) <yi.zhang@xxxxxxxxxx> >>>> --- >>>> README.overlay | 10 ++++-- >>>> common/config | 4 +++ >>>> common/overlay | 99 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ >>>> common/rc | 18 +++++++++-- >>>> 4 files changed, 127 insertions(+), 4 deletions(-) >>>> >>>> diff --git a/README.overlay b/README.overlay >>>> index dfb8234..3cbefab 100644 >>>> --- a/README.overlay >>>> +++ b/README.overlay >>>> @@ -22,6 +22,10 @@ the base fs should be pre-formatted before starting the -overlay run. >>>> An easy way to accomplish this is by running './check <some test>' once, >>>> before running './check -overlay'. >>>> >>>> +'./check -overlay' support check overlay test and scratch dirs, >>>> +OVERLAY_FSCK_OPTIONS should be set instead of FSCK_OPTIONS if fsck >>>> +options need to given directly. >>>> + >>>> Because of the lack of mkfs support, multi-section config files are only >>>> partly supported with './check -overlay'. Only multi-section files that >>>> do not change FSTYP and MKFS_OPTIONS can be safely used with -overlay. >>>> @@ -40,7 +44,9 @@ run overlay tests on the same base fs, but with different mount options: >>>> MOUNT_OPTIONS="-o pquota" >>>> TEST_FS_MOUNT_OPTS="-o noatime" >>>> OVERLAY_MOUNT_OPTIONS="-o redirect_dir=off" >>>> + OVERLAY_FSCK_OPTIONS="-n redirect_dir=off" >>> >>> typo for the phony options. I believe it is going to be: "-n -o >>> redirect_dir=off" >> >> Oh, sorry for the mistake, it is "-n -o redirect_dir=off" >> > > Zhangyi, > > One thing that has occurred to me now, which should be fixed before > first release > is that fsck.overlay should return an error for unknown/unsupported options > (e.g. fsck.overlay -o blah doesn't complain). > If this doesn't happen for first release then in the future we will > not be able to > write test scripts to check if installed fsck.overlay version supports > feature X. > Yes, It's time to impliment the feature/option set and do the corresponding work for the already known features. I am writing the kernel side feature set recent days, I will add these feature into fsck.overlay after kernel's work done. BTW, the feature set in kernel do the following, some suggestions? 1) Add feature set into upper root xattr "trusted.overlay.feature_[compat|imcompat|rocompat]" if user give mount options "xxx=on" or some feature xattr/dir detected. 2) If unknown imcompat feature in xattr detected, fall back to try to ro-mount. 3) If unknown ro-compat featrure in xattr detected, mount fail. > About the possible meaning of <feature>=on/off and the meaning of <defaults> > in the context of fsck.overlay, I think it is worth a separate > discussion, but the > way I see it: > - When fsck.overlay *supports* a <feature> it should be able to detect if that > <feature> was ever enabled on overlay (e.g. known xattr exists or > index dir exists). Agree, and we can also fix feature set when detect feature. > - When fsck.overlay detects an unknown <feature> it should abort Agree. A little more detail: if user specify upper and work dir, check imcompat unknown feature; if not, check ro-imcompat unknown feature. If detected, fsck about. > - -o <feature>=off means wipe all traces of <feature> from overlay and fix if > necessary (e.g. make a copy of redirect dir before removing > redirect_dir xattr) Agree. But not sure it is easy to implement of making a copy of redirect dir when redirect_dir=off. Because it modify the directory hierarchy, it may affect the origin/nlink xattr in files/dirs in index dir when index=on or nfs_export=on, maybe also affect some other feature future. > - -o <feature>=on means bring overlay to a "feature consistent state", > as if feature > was enabled from day 1 (e.g. index all copies up lower hardlinks) Agree. > - If <feature>=[on/off] is not specified, then if <feature> is not > detected, it should > not be explicitly enabled and if feature is detected it should be > made consistent Agree. > That last rule is certainly debatable, so maybe best if we leave that > decision to > admin via default policy configuration file. > Something like /etc/mke2fs.conf but not for mkfs? Do you mean if feature detected conflict with this config(eg: we detect redirect dir but "redirect_dir=off" was setted in ths config file), fsck should check and fix fs according to this config file? I think it may confusion and conflict with mount options(eg: redirect_dir=on) previous mount or next mount. And IIUC, there is no such feature config file for ext4/xfs, fsck.ext4 will detect feature from super block and fix fs according to this detected result. So I think it's better to detect feature feature by fsck itself, keep features as it was. Thanks, Yi. -- To unsubscribe from this list: send the line "unsubscribe linux-unionfs" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
