| Message ID | 20220827002815.19116-1-pvorel@suse.cz |
|---|---|
| Headers | show |
| Series | Track minimal size per filesystem | expand |
Hi! > This patchset require to be on the top of: > > [RFC,1/1] API: Allow to use xfs filesystems < 300 MB > https://lore.kernel.org/ltp/20220817204015.31420-1-pvorel@suse.cz/ > https://patchwork.ozlabs.org/project/ltp/patch/20220817204015.31420-1-pvorel@suse.cz/ I'm not that sure if we want to run tests for xfs filesystem that is smaller than minimal size used in production. I bet that we will cover different codepaths that eventually end up being used in production that way.
Hi Cyril, > Hi! > > This patchset require to be on the top of: > > [RFC,1/1] API: Allow to use xfs filesystems < 300 MB > > https://lore.kernel.org/ltp/20220817204015.31420-1-pvorel@suse.cz/ > > https://patchwork.ozlabs.org/project/ltp/patch/20220817204015.31420-1-pvorel@suse.cz/ > I'm not that sure if we want to run tests for xfs filesystem that is > smaller than minimal size used in production. I bet that we will cover > different codepaths that eventually end up being used in production > that way. > > LTP community: do we want to depend on this behavior or we just increase from 256MB to 301 MB > > (either for XFS or for all). It might not be a good idea to test size users are required > > to use. > It might *not*? <confused> Again, I'm sorry, missing another not. I.e. I suppose normal users will not try to go below 301MB, therefore LTP probably should not do it either. That's why RFC. @Darrick, others (kernel/LTP maintainers, embedded folks) WDYT? I'm personally OK to use 300 MB (safer to use code paths which are used in production), it's just that for older kernels even with xfs-progs installed it's unnecessary boundary. We could base XFS size on runtime kernel, but unless it's 300 MB a real problem for anybody I would not address it. i.e. is there anybody using XFS on old kernels? (old LTS, whey sooner or later need to use these variables themselves). Kind regards, Petr [1] https://lore.kernel.org/ltp/Yv4ABHlsP+BZ3bRD@pevik/
Hi Petr, All, On Mon, Aug 29, 2022 at 7:40 PM Petr Vorel <pvorel@suse.cz> wrote: > Hi Cyril, > > > Hi! > > > This patchset require to be on the top of: > > > > [RFC,1/1] API: Allow to use xfs filesystems < 300 MB > > > https://lore.kernel.org/ltp/20220817204015.31420-1-pvorel@suse.cz/ > > > > https://patchwork.ozlabs.org/project/ltp/patch/20220817204015.31420-1-pvorel@suse.cz/ > > > I'm not that sure if we want to run tests for xfs filesystem that is > > smaller than minimal size used in production. I bet that we will cover > > different codepaths that eventually end up being used in production > > that way. > > > > LTP community: do we want to depend on this behavior or we > just increase from 256MB to 301 MB > > > (either for XFS or for all). It might not be a good idea to > test size users are required > > > to use. > > > It might *not*? <confused> > Again, I'm sorry, missing another not. I.e. I suppose normal users > will not try > to go below 301MB, therefore LTP probably should not do it either. > That's why > RFC. > > @Darrick, others (kernel/LTP maintainers, embedded folks) WDYT? > > I'm personally OK to use 300 MB (safer to use code paths which are used in > production), it's just that for older kernels even with xfs-progs > installed it's > unnecessary boundary. We could base XFS size on runtime kernel, but unless > it's > 300 MB a real problem for anybody I would not address it. i.e. is there > anybody > using XFS on old kernels? (old LTS, whey sooner or later need to use these > variables themselves). > Another compromised way I can think of is to let LTP choose 300MB for XFS by default, if the test bed can't provide that size, simply go back to try 16MB. Does this sound acceptable?
> Hi Petr, All, > On Mon, Aug 29, 2022 at 7:40 PM Petr Vorel <pvorel@suse.cz> wrote: > > Hi Cyril, > > > Hi! > > > > This patchset require to be on the top of: > > > > [RFC,1/1] API: Allow to use xfs filesystems < 300 MB > > > > https://lore.kernel.org/ltp/20220817204015.31420-1-pvorel@suse.cz/ > > https://patchwork.ozlabs.org/project/ltp/patch/20220817204015.31420-1-pvorel@suse.cz/ > > > I'm not that sure if we want to run tests for xfs filesystem that is > > > smaller than minimal size used in production. I bet that we will cover > > > different codepaths that eventually end up being used in production > > > that way. > > > > LTP community: do we want to depend on this behavior or we > > just increase from 256MB to 301 MB > > > > (either for XFS or for all). It might not be a good idea to > > test size users are required > > > > to use. > > > It might *not*? <confused> > > Again, I'm sorry, missing another not. I.e. I suppose normal users > > will not try > > to go below 301MB, therefore LTP probably should not do it either. > > That's why > > RFC. > > @Darrick, others (kernel/LTP maintainers, embedded folks) WDYT? > > I'm personally OK to use 300 MB (safer to use code paths which are used in > > production), it's just that for older kernels even with xfs-progs > > installed it's > > unnecessary boundary. We could base XFS size on runtime kernel, but unless > > it's > > 300 MB a real problem for anybody I would not address it. i.e. is there > > anybody > > using XFS on old kernels? (old LTS, whey sooner or later need to use these > > variables themselves). > Another compromised way I can think of is to let LTP choose > 300MB for XFS by default, if the test bed can't provide that size, > simply go back to try 16MB. Does this sound acceptable? I'll try to have look into this, but it'd would be quite special case given we don't try to detect and recovery mkfs.* failures. Kind regards, Petr