| Message ID | 20240125004732.9549-1-wegao@suse.com |
|---|---|
| State | Accepted |
| Headers | show |
| Series | [v1] fs_fill: Increase test loop device size to 1GB | expand |
Hi Wei, > On PPC64 the page size is 64K and this causes trouble on btrfs nit: s/PPC64/ppc64le/ (the other one is 32 bit) nit: s/64K/4K/ > filesystems of small size(LTP currently use 300M), the threads > could compete for a very small number of pages/blocks to actually > write the data. So Increase minimal device size to 1G avoid the > corner case. Reviewed-by: Petr Vorel <pvorel@suse.cz> I wonder if would work with less (512 MB), but probably better to be safe with 1024 MB. Kind regards, Petr
On 25. 01. 24 10:35, Petr Vorel wrote: > Hi Wei, > >> On PPC64 the page size is 64K and this causes trouble on btrfs > nit: s/PPC64/ppc64le/ (the other one is 32 bit) The other one is 64bit but big-endian. And yes, the big-endian mode will have the same issue, so "PPC64" is good enough catch-all term. > nit: s/64K/4K/ No. >> filesystems of small size(LTP currently use 300M), the threads >> could compete for a very small number of pages/blocks to actually >> write the data. So Increase minimal device size to 1G avoid the >> corner case. > > Reviewed-by: Petr Vorel <pvorel@suse.cz> > > I wonder if would work with less (512 MB), but probably better to be safe with > 1024 MB. > > Kind regards, > Petr >
Hi, Reviewed-by: Martin Doucha <mdoucha@suse.cz> On 25. 01. 24 1:47, Wei Gao via ltp wrote: > On PPC64 the page size is 64K and this causes trouble on btrfs > filesystems of small size(LTP currently use 300M), the threads > could compete for a very small number of pages/blocks to actually > write the data. So Increase minimal device size to 1G avoid the > corner case. > > Signed-off-by: Wei Gao <wegao@suse.com> > --- > testcases/kernel/fs/fs_fill/fs_fill.c | 1 + > 1 file changed, 1 insertion(+) > > diff --git a/testcases/kernel/fs/fs_fill/fs_fill.c b/testcases/kernel/fs/fs_fill/fs_fill.c > index 2ecd8e2ad..325e83643 100644 > --- a/testcases/kernel/fs/fs_fill/fs_fill.c > +++ b/testcases/kernel/fs/fs_fill/fs_fill.c > @@ -123,6 +123,7 @@ static void cleanup(void) > static struct tst_test test = { > .max_runtime = 60, > .needs_root = 1, > + .dev_min_size = 1024, > .mount_device = 1, > .mntpoint = MNTPOINT, > .all_filesystems = 1,
> On 25. 01. 24 10:35, Petr Vorel wrote: > > Hi Wei, > > > On PPC64 the page size is 64K and this causes trouble on btrfs > > nit: s/PPC64/ppc64le/ (the other one is 32 bit) > The other one is 64bit but big-endian. And yes, the big-endian mode will > have the same issue, so "PPC64" is good enough catch-all term. Wei, I'm sorry, of course Martin, you're right. > > nit: s/64K/4K/ > No. Also here. Martin, thanks for correcting me. Kind regards, Petr > > > filesystems of small size(LTP currently use 300M), the threads > > > could compete for a very small number of pages/blocks to actually > > > write the data. So Increase minimal device size to 1G avoid the > > > corner case. > > Reviewed-by: Petr Vorel <pvorel@suse.cz> > > I wonder if would work with less (512 MB), but probably better to be safe with > > 1024 MB. > > Kind regards, > > Petr
diff --git a/testcases/kernel/fs/fs_fill/fs_fill.c b/testcases/kernel/fs/fs_fill/fs_fill.c index 2ecd8e2ad..325e83643 100644 --- a/testcases/kernel/fs/fs_fill/fs_fill.c +++ b/testcases/kernel/fs/fs_fill/fs_fill.c @@ -123,6 +123,7 @@ static void cleanup(void) static struct tst_test test = { .max_runtime = 60, .needs_root = 1, + .dev_min_size = 1024, .mount_device = 1, .mntpoint = MNTPOINT, .all_filesystems = 1,
On PPC64 the page size is 64K and this causes trouble on btrfs filesystems of small size(LTP currently use 300M), the threads could compete for a very small number of pages/blocks to actually write the data. So Increase minimal device size to 1G avoid the corner case. Signed-off-by: Wei Gao <wegao@suse.com> --- testcases/kernel/fs/fs_fill/fs_fill.c | 1 + 1 file changed, 1 insertion(+)