| Message ID | 20210612123154.8098-1-jlayton@kernel.org (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | generic: add a test to ensure that page is properly filled before write | expand |
On Sat, Jun 12, 2021 at 08:31:54AM -0400, Jeff Layton wrote: > We had a broken optimization in cephfs and netfs lib that could cause > part of a page to be improperly zeroed-out when writing to an offset > that was beyond the EOF but in an existing page. > > Add a simple test that would have caught this. > > Signed-off-by: Jeff Layton <jlayton@kernel.org> > --- > tests/generic/XXX | 53 +++++++++++++++++++++++++++++++++++++++++++ > tests/generic/XXX.out | 5 ++++ Ah, all the 'XXX' in this test should be replaced with proper seq number. > tests/generic/group | 1 + > 3 files changed, 59 insertions(+) > create mode 100755 tests/generic/XXX > create mode 100644 tests/generic/XXX.out > > diff --git a/tests/generic/XXX b/tests/generic/XXX > new file mode 100755 > index 000000000000..0ddaaa544609 > --- /dev/null > +++ b/tests/generic/XXX > @@ -0,0 +1,53 @@ > +#! /bin/bash > +# SPDX-License-Identifier: GPL-2.0 > +# Copyright (c) 2021, Jeff Layton <jlayton@redhat.com> > +# > +# FS QA Test No. XXX > +# > +# Open a file and write a little data to it. Unmount (to clean out the cache) > +# and then mount again. Then write some data to it beyond the EOF and ensure > +# the result is correct. Prompted by a bug in ceph_write_begin. See: > +# > +# https://lore.kernel.org/ceph-devel/97002.1623448034@warthog.procyon.org.uk/T/#m9f34b328623b493c529505e12f64f4dfdbddfb27 This looks like a thread 9 months ago. Was the fix merged? If so, it'd be better to reference the kernel commit here. Otherwise test looks fine to me. Thanks, Eryu > +# > +seq=`basename $0` > +seqres=$RESULT_DIR/$seq > +echo "QA output created by $seq" > + > +here=`pwd` > +status=1 # failure is the default! > +trap "_cleanup; exit \$status" 0 1 2 3 15 > + > + > +_cleanup() > +{ > + cd / > + rm -f $testfile > +} > + > +# get standard environment, filters and checks > +. ./common/rc > +. ./common/filter > + > +_supported_fs generic > +_require_test > + > +rm -f $seqres.full > + > +testfile="$TEST_DIR/test_write_begin.$$" > + > +# write some data to file and fsync it out > +$XFS_IO_PROG -f -c "pwrite -q 0 32" $testfile > + > +# cycle the mount to clean out the pagecache > +_test_cycle_mount > + > +# now, write to the file (near the end) > +$XFS_IO_PROG -c "pwrite -q 32 32" $testfile > + > +# dump what we think is in there > +echo "The result should be 64 bytes filled with 0xcd:" > +hexdump -C $testfile > + > +status=0 > +exit > diff --git a/tests/generic/XXX.out b/tests/generic/XXX.out > new file mode 100644 > index 000000000000..4e7653858e51 > --- /dev/null > +++ b/tests/generic/XXX.out > @@ -0,0 +1,5 @@ > +QA output created by XXX > +The result should be 64 bytes filled with 0xcd: > +00000000 cd cd cd cd cd cd cd cd cd cd cd cd cd cd cd cd |................| > +* > +00000040 > diff --git a/tests/generic/group b/tests/generic/group > index 9a636b23f243..204a9b548f78 100644 > --- a/tests/generic/group > +++ b/tests/generic/group > @@ -641,3 +641,4 @@ > 636 auto quick swap > 637 auto quick dir > 638 auto quick rw > +XXX auto quick rw > -- > 2.31.1 >
On Sun, 2021-06-13 at 22:52 +0800, Eryu Guan wrote: > On Sat, Jun 12, 2021 at 08:31:54AM -0400, Jeff Layton wrote: > > We had a broken optimization in cephfs and netfs lib that could cause > > part of a page to be improperly zeroed-out when writing to an offset > > that was beyond the EOF but in an existing page. > > > > Add a simple test that would have caught this. > > > > Signed-off-by: Jeff Layton <jlayton@kernel.org> > > --- > > tests/generic/XXX | 53 +++++++++++++++++++++++++++++++++++++++++++ > > tests/generic/XXX.out | 5 ++++ > > Ah, all the 'XXX' in this test should be replaced with proper seq > number. > I can do that and resend. > > tests/generic/group | 1 + > > 3 files changed, 59 insertions(+) > > create mode 100755 tests/generic/XXX > > create mode 100644 tests/generic/XXX.out > > > > diff --git a/tests/generic/XXX b/tests/generic/XXX > > new file mode 100755 > > index 000000000000..0ddaaa544609 > > --- /dev/null > > +++ b/tests/generic/XXX > > @@ -0,0 +1,53 @@ > > +#! /bin/bash > > +# SPDX-License-Identifier: GPL-2.0 > > +# Copyright (c) 2021, Jeff Layton <jlayton@redhat.com> > > +# > > +# FS QA Test No. XXX > > +# > > +# Open a file and write a little data to it. Unmount (to clean out the cache) > > +# and then mount again. Then write some data to it beyond the EOF and ensure > > +# the result is correct. Prompted by a bug in ceph_write_begin. See: > > +# > > +# https://lore.kernel.org/ceph-devel/97002.1623448034@warthog.procyon.org.uk/T/#m9f34b328623b493c529505e12f64f4dfdbddfb27 > > This looks like a thread 9 months ago. Was the fix merged? If so, it'd > be better to reference the kernel commit here. > > Otherwise test looks fine to me. > Not yet. v5.13 will require a fix in the netfs lib, whereas ceph will need a one-off fix in v5.10.z - v5.12.z. We're sorting out both fixes now. Thanks! > > > +# > > +seq=`basename $0` > > +seqres=$RESULT_DIR/$seq > > +echo "QA output created by $seq" > > + > > +here=`pwd` > > +status=1 # failure is the default! > > +trap "_cleanup; exit \$status" 0 1 2 3 15 > > + > > + > > +_cleanup() > > +{ > > + cd / > > + rm -f $testfile > > +} > > + > > +# get standard environment, filters and checks > > +. ./common/rc > > +. ./common/filter > > + > > +_supported_fs generic > > +_require_test > > + > > +rm -f $seqres.full > > + > > +testfile="$TEST_DIR/test_write_begin.$$" > > + > > +# write some data to file and fsync it out > > +$XFS_IO_PROG -f -c "pwrite -q 0 32" $testfile > > + > > +# cycle the mount to clean out the pagecache > > +_test_cycle_mount > > + > > +# now, write to the file (near the end) > > +$XFS_IO_PROG -c "pwrite -q 32 32" $testfile > > + > > +# dump what we think is in there > > +echo "The result should be 64 bytes filled with 0xcd:" > > +hexdump -C $testfile > > + > > +status=0 > > +exit > > diff --git a/tests/generic/XXX.out b/tests/generic/XXX.out > > new file mode 100644 > > index 000000000000..4e7653858e51 > > --- /dev/null > > +++ b/tests/generic/XXX.out > > @@ -0,0 +1,5 @@ > > +QA output created by XXX > > +The result should be 64 bytes filled with 0xcd: > > +00000000 cd cd cd cd cd cd cd cd cd cd cd cd cd cd cd cd |................| > > +* > > +00000040 > > diff --git a/tests/generic/group b/tests/generic/group > > index 9a636b23f243..204a9b548f78 100644 > > --- a/tests/generic/group > > +++ b/tests/generic/group > > @@ -641,3 +641,4 @@ > > 636 auto quick swap > > 637 auto quick dir > > 638 auto quick rw > > +XXX auto quick rw > > -- > > 2.31.1 > >
diff --git a/tests/generic/XXX b/tests/generic/XXX new file mode 100755 index 000000000000..0ddaaa544609 --- /dev/null +++ b/tests/generic/XXX @@ -0,0 +1,53 @@ +#! /bin/bash +# SPDX-License-Identifier: GPL-2.0 +# Copyright (c) 2021, Jeff Layton <jlayton@redhat.com> +# +# FS QA Test No. XXX +# +# Open a file and write a little data to it. Unmount (to clean out the cache) +# and then mount again. Then write some data to it beyond the EOF and ensure +# the result is correct. Prompted by a bug in ceph_write_begin. See: +# +# https://lore.kernel.org/ceph-devel/97002.1623448034@warthog.procyon.org.uk/T/#m9f34b328623b493c529505e12f64f4dfdbddfb27 +# +seq=`basename $0` +seqres=$RESULT_DIR/$seq +echo "QA output created by $seq" + +here=`pwd` +status=1 # failure is the default! +trap "_cleanup; exit \$status" 0 1 2 3 15 + + +_cleanup() +{ + cd / + rm -f $testfile +} + +# get standard environment, filters and checks +. ./common/rc +. ./common/filter + +_supported_fs generic +_require_test + +rm -f $seqres.full + +testfile="$TEST_DIR/test_write_begin.$$" + +# write some data to file and fsync it out +$XFS_IO_PROG -f -c "pwrite -q 0 32" $testfile + +# cycle the mount to clean out the pagecache +_test_cycle_mount + +# now, write to the file (near the end) +$XFS_IO_PROG -c "pwrite -q 32 32" $testfile + +# dump what we think is in there +echo "The result should be 64 bytes filled with 0xcd:" +hexdump -C $testfile + +status=0 +exit diff --git a/tests/generic/XXX.out b/tests/generic/XXX.out new file mode 100644 index 000000000000..4e7653858e51 --- /dev/null +++ b/tests/generic/XXX.out @@ -0,0 +1,5 @@ +QA output created by XXX +The result should be 64 bytes filled with 0xcd: +00000000 cd cd cd cd cd cd cd cd cd cd cd cd cd cd cd cd |................| +* +00000040 diff --git a/tests/generic/group b/tests/generic/group index 9a636b23f243..204a9b548f78 100644 --- a/tests/generic/group +++ b/tests/generic/group @@ -641,3 +641,4 @@ 636 auto quick swap 637 auto quick dir 638 auto quick rw +XXX auto quick rw
We had a broken optimization in cephfs and netfs lib that could cause part of a page to be improperly zeroed-out when writing to an offset that was beyond the EOF but in an existing page. Add a simple test that would have caught this. Signed-off-by: Jeff Layton <jlayton@kernel.org> --- tests/generic/XXX | 53 +++++++++++++++++++++++++++++++++++++++++++ tests/generic/XXX.out | 5 ++++ tests/generic/group | 1 + 3 files changed, 59 insertions(+) create mode 100755 tests/generic/XXX create mode 100644 tests/generic/XXX.out