Tests get stuck. #261
Description
I was trying to run go-fuse testsuite in repeated mode and after some 10-20-30 iterations the tests get stuck:
src/github.com/hanwen/go-fuse$ while true; do echo -e '\n# repeat'; go test -count 1 ./...; done
...
# repeat
ok github.com/hanwen/go-fuse/benchmark 0.058s
? github.com/hanwen/go-fuse/benchmark/bulkstat [no test files]
? github.com/hanwen/go-fuse/example/autounionfs [no test files]
? github.com/hanwen/go-fuse/example/benchmark-read-throughput [no test files]
? github.com/hanwen/go-fuse/example/hello [no test files]
? github.com/hanwen/go-fuse/example/loopback [no test files]
? github.com/hanwen/go-fuse/example/memfs [no test files]
? github.com/hanwen/go-fuse/example/multizip [no test files]
? github.com/hanwen/go-fuse/example/statfs [no test files]
? github.com/hanwen/go-fuse/example/unionfs [no test files]
? github.com/hanwen/go-fuse/example/zipfs [no test files]
ok github.com/hanwen/go-fuse/fuse 0.002s
ok github.com/hanwen/go-fuse/fuse/nodefs 0.235s
ok github.com/hanwen/go-fuse/fuse/pathfs 0.207s
ok github.com/hanwen/go-fuse/fuse/test 3.037s
? github.com/hanwen/go-fuse/internal/testutil [no test files]
? github.com/hanwen/go-fuse/internal/utimens [no test files]
ok github.com/hanwen/go-fuse/splice 0.011s
ok github.com/hanwen/go-fuse/unionfs 2.892s
ok github.com/hanwen/go-fuse/zipfs 0.114s
# repeat
ok github.com/hanwen/go-fuse/benchmark 0.047s
? github.com/hanwen/go-fuse/benchmark/bulkstat [no test files]
? github.com/hanwen/go-fuse/example/autounionfs [no test files]
? github.com/hanwen/go-fuse/example/benchmark-read-throughput [no test files]
? github.com/hanwen/go-fuse/example/hello [no test files]
? github.com/hanwen/go-fuse/example/loopback [no test files]
? github.com/hanwen/go-fuse/example/memfs [no test files]
? github.com/hanwen/go-fuse/example/multizip [no test files]
? github.com/hanwen/go-fuse/example/statfs [no test files]
? github.com/hanwen/go-fuse/example/unionfs [no test files]
? github.com/hanwen/go-fuse/example/zipfs [no test files]
ok github.com/hanwen/go-fuse/fuse 0.003s
ok github.com/hanwen/go-fuse/fuse/nodefs 0.242s
ok github.com/hanwen/go-fuse/fuse/pathfs 0.292s
(no further output here)
inspecting the stuck process a bit:
kirr@deco:~$ cat /proc/12927/mounts |grep -i fuse
fusectl /sys/fs/fuse/connections fusectl rw,relatime 0 0
nodefs.memNode /tmp/TestDeleteNotify043745505/mnt fuse.nodefs.memNode rw,nosuid,nodev,relatime,user_id=1000,group_id=1000 0 0
nodefs.memNode /tmp/TestDeleteNotify041170318/mnt fuse.nodefs.memNode rw,nosuid,nodev,relatime,user_id=1000,group_id=1000 0 0
nodefs.memNode /tmp/TestDeleteNotify074247721/mnt fuse.nodefs.memNode rw,nosuid,nodev,relatime,user_id=1000,group_id=1000 0 0
pathfs.pathInode /tmp/TestNonseekable501982173 fuse.pathfs.pathInode rw,nosuid,nodev,relatime,user_id=1000,group_id=1000 0 0
nodefs.memNode /tmp/TestDeleteNotify426583496/mnt fuse.nodefs.memNode rw,nosuid,nodev,relatime,user_id=1000,group_id=1000 0 0
Locked(pathfs.pathIn /tmp/TestRecursiveMount699955034/mnt fuse.Locked(pathfs.pathIn rw,nosuid,nodev,relatime,user_id=1000,group_id=1000 0 0
Locked(pathfs.pathIn /tmp/TestLookupKnownChildrenAttrCopied546363126/mnt fuse.Locked(pathfs.pathIn rw,nosuid,nodev,relatime,user_id=1000,group_id=1000 0 0
nodefs.memNode /tmp/TestDeleteNotify591838373/mnt fuse.nodefs.memNode rw,nosuid,nodev,relatime,user_id=1000,group_id=1000 0 0
pathfs.pathInode /tmp/TestUnionFsDropDeletionCache497637161/mnt fuse.pathfs.pathInode rw,nosuid,nodev,relatime,user_id=1000,group_id=1000 0 0
Locked(pathfs.pathIn /tmp/TestRecursiveMount577527402/mnt fuse.Locked(pathfs.pathIn rw,nosuid,nodev,relatime,user_id=1000,group_id=1000 0 0
pathfs.pathInode /tmp/TestUnionFsRenameDirAllSourcesGone103878418/mnt fuse.pathfs.pathInode rw,nosuid,nodev,relatime,user_id=1000,group_id=1000 0 0
nodefs.nodeReadNode /tmp/nodefs272940596 fuse.nodefs.nodeReadNode rw,nosuid,nodev,relatime,user_id=1000,group_id=1000 0 0
nodefs.memNode /tmp/TestDeleteNotify228516545/mnt fuse.nodefs.memNode rw,nosuid,nodev,relatime,user_id=1000,group_id=1000 0 0
kirr@deco:~$ ll /proc/12927/task/
итого 0
dr-xr-xr-x 7 kirr kirr 0 янв 31 11:42 12927
i.e. single thread.
kirr@deco:~$ strace -t -t -p 12927
strace: Process 12927 attached
(no further output)
i.e. it is blocked in a syscall. Syscall details:
root@deco:/home/kirr# cat /proc/12927/stack
[<0>] request_wait_answer+0x12b/0x210 [fuse]
[<0>] __fuse_request_send+0x64/0x90 [fuse]
[<0>] fuse_simple_request+0xcd/0x190 [fuse]
[<0>] fuse_dentry_revalidate+0x1a8/0x300 [fuse]
[<0>] lookup_fast+0x275/0x2e0
[<0>] walk_component+0x48/0x4a0
[<0>] path_lookupat.isra.48+0x6d/0x220
[<0>] filename_lookup.part.62+0xa0/0x170
[<0>] ksys_chdir+0x3e/0xc0
[<0>] __x64_sys_chdir+0xe/0x20
[<0>] do_syscall_64+0x53/0x100
[<0>] entry_SYSCALL_64_after_hwframe+0x44/0xa9
[<0>] 0xffffffffffffffff
I had only a brief quick look and had not inspected further.