Client hangs

[arbee]

Hi

I'm trying to pin down the cause of some beegfs client hangs that cause processes accessing beegfs to completely lock up in kernel space.

Some of the messages reported are:

myhostname kernel: beegfs: pager(3128572): Messaging (RPC): Communication error: Node: beegfs-meta pod9b.loc [ID: 102] (comm result: -13; message type: 2059)

When the kernel reports a hung task it gives a stack trace like this:

kernel: Call Trace:

kernel:  __schedule+0x2e3/0x740

kernel:  schedule+0x42/0xb0

kernel:  rwsem_down_write_slowpath+0x244/0x4d0

kernel:  ? ktime_get_ts64+0x4c/0xe0

kernel:  down_write+0x41/0x50

kernel:  __FhgfsOps_revalidateIntent.isra.0+0x118/0x4a0 [beegfs]

kernel:  ? xas_load+0xd/0x80

kernel:  ? find_get_entry+0xaf/0x170

kernel:  ? pagecache_get_page+0x15b/0x300

kernel:  ? __find_get_block+0xb9/0x2d0

kernel:  ? __check_block_validity.constprop.0+0x43/0x90

kernel:  FhgfsOps_revalidateIntent+0xc4/0x110 [beegfs]

kernel:  lookup_fast+0x281/0x300

kernel:  walk_component+0x48/0x360

kernel:  ? inode_permission+0x65/0x1b0

kernel:  link_path_walk.part.0+0x2cc/0x550

kernel:  path_lookupat.isra.0+0x4e/0x230

kernel:  ? _cond_resched+0x19/0x30

kernel:  ? down_read+0x13/0xa0

kernel:  filename_lookup+0xae/0x170

kernel:  ? __check_object_size+0x13f/0x150

kernel:  ? strncpy_from_user+0x4c/0x150

kernel:  user_path_at_empty+0x3a/0x50

kernel:  vfs_statx+0x7d/0xe0

kernel:  ? prepend_path.isra.0+0xf4/0x300

kernel:  __do_sys_newstat+0x3e/0x80

kernel:  ? __x64_sys_getcwd+0x123/0x1f0

kernel:  __x64_sys_newstat+0x16/0x20

kernel:  do_syscall_64+0x57/0x190

kernel:  entry_SYSCALL_64_after_hwframe+0x44/0xa9

This is on ubuntu 20.04 beegfs 7.2

Thanks