On 14/05/15 03:53, Sébastien Luttringer wrote: > Hello, > > I experience one test suite failure when I tried to rebuild the > coreutils v8.23 package. My system is an archlinux host with a btrfs > filesystem. > > The test doesn't fail when I run it in a tmpfs. I'm pretty sure it used > to works over btrfs in july 2014[1] when I built the previous package > for archlinux. > > The failing section is the following: > > # Ensure NUL blocks smaller than the block size are not made sparse. > # Here, with a 2MiB block size, dd's conv=sparse must *not* > introduce a hole. > dd if=file.in of=file.out bs=2M conv=sparse > test 2500 -lt $(kb_alloc file.out) || fail=1 > > You can see the full shell output here: http://pastebin.com/XADb1BKj > > Regards, > > [1] https://www.archlinux.org/packages/core/x86_64/coreutils/
It seems the allocation is async. I also see this intermittently with BTRFS on 4.0.0-0.rc5.git4.1.fc22.x86_64 I've been unable to reproduce with the attached which syncs the file before checking the allocation. thanks! Pádraig.
From 77e092c85ed9ae2221042323cf87f4fc8c847e75 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?P=C3=A1draig=20Brady?= <[email protected]> Date: Thu, 14 May 2015 11:43:54 +0100 Subject: [PATCH] tests: fix async allocation race on BTRFS * tests/dd/sparse.sh: Sync files before checking allocations, which may be done asynchronously on NFS and BTRFS at least. Also mark this test as very expensive on remote file systems. * tests/tail-2/inotify-rotate-resources.sh: Use the more standard is_local_dir_() to check remoteness. * tests/cp/fiemap-empty.sh: Comment on the sync issue for this currerntly unused test. Fixes http://bugs.gnu.org/20570 --- tests/cp/fiemap-empty.sh | 1 + tests/dd/sparse.sh | 9 ++++++++- tests/tail-2/inotify-rotate-resources.sh | 2 +- 3 files changed, 10 insertions(+), 2 deletions(-) diff --git a/tests/cp/fiemap-empty.sh b/tests/cp/fiemap-empty.sh index a5eaac5..b3b2cd7 100755 --- a/tests/cp/fiemap-empty.sh +++ b/tests/cp/fiemap-empty.sh @@ -22,6 +22,7 @@ print_ver_ cp # FIXME: enable any part of this test that is still relevant, # or, if none are relevant (now that cp does not handle unwritten # extents), just remove the test altogether. +# Note also if checking allocations may need to sync first on BTRFS at least skip_ 'disabled for now' touch fiemap_chk diff --git a/tests/dd/sparse.sh b/tests/dd/sparse.sh index 4fdabad..fde5b32 100755 --- a/tests/dd/sparse.sh +++ b/tests/dd/sparse.sh @@ -17,6 +17,7 @@ . "${srcdir=.}/tests/init.sh"; path_prepend_ ./src print_ver_ dd +is_local_dir_ . || very_expensive_ require_sparse_support_ # Ensure basic sparse generation works @@ -50,6 +51,9 @@ dd if=/dev/zero of=file.in bs=1M count=1 seek=1 conv=notrunc || fail=1 kb_alloc() { du -k "$1"|cut -f1; } +# sync out data for async allocators like NFS/BTRFS +# sync file.in || fail=1 + # If our just-created input file appears to be too small, # skip the remaining tests. On at least Solaris 10 with NFS, # file.in is reported to occupy <= 1KiB for about 50 seconds @@ -58,7 +62,10 @@ if test $(kb_alloc file.in) -gt 3000; then # Ensure NUL blocks smaller than the block size are not made sparse. # Here, with a 2MiB block size, dd's conv=sparse must *not* introduce a hole. - dd if=file.in of=file.out bs=2M conv=sparse + dd if=file.in of=file.out bs=2M conv=sparse || fail=1 + + # Intermittently BTRFS returns 0 allocation for file.out unless synced + sync file.out || fail=1 test 2500 -lt $(kb_alloc file.out) || fail=1 # Note we recreate a sparse file first to avoid diff --git a/tests/tail-2/inotify-rotate-resources.sh b/tests/tail-2/inotify-rotate-resources.sh index c13959e..d6629a8 100755 --- a/tests/tail-2/inotify-rotate-resources.sh +++ b/tests/tail-2/inotify-rotate-resources.sh @@ -25,7 +25,7 @@ grep '^#define HAVE_INOTIFY 1' "$CONFIG_HEADER" >/dev/null \ require_strace_ 'inotify_add_watch,inotify_rm_watch' # Quickly skip on remote file systems -df --local . >/dev/null 2>&1 || +is_local_dir . >/dev/null 2>&1 || skip_ 'inotify not used on remote file system' check_tail_output() -- 2.3.4
