linux-stable.git/fs/dlm, branch linux-6.15.y

dlm: use SHUT_RDWR for SCTP shutdown

2025-06-27T10:13:16+00:00

[ Upstream commit 55612ddb62fc12437a7ff2f27b51a8981bc187a4 ]

Currently SCTP shutdown() call gets stuck because there is no incoming
EOF indicator on its socket. On the peer side the EOF indicator as
recvmsg() returns 0 will be triggered as mechanism to flush the socket
queue on the receive side. In SCTP recvmsg() function sctp_recvmsg() we
can see that only if sk_shutdown has the bit RCV_SHUTDOWN set SCTP will
recvmsg() will return EOF. The RCV_SHUTDOWN bit will only be set when
shutdown with SHUT_RD is called. We use now SHUT_RDWR to also get a EOF
indicator from recvmsg() call on the shutdown() initiator.

SCTP does not support half closed sockets and the semantic of SHUT_WR is
different here, it seems that calling SHUT_WR on sctp sockets keeps the
socket open to have the possibility to do some specific SCTP operations on
it that we don't do here.

There exists still a difference in the limitations of TCP vs SCTP in
case if we are required to have a half closed socket functionality. This
was tried to archieve with DLM protocol changes in the past and
hopefully we really don't require half closed socket functionality.

Signed-off-by: Alexander Aring 
Tested-by: Heming zhao 
Reviewed-by: Heming zhao 
Signed-off-by: David Teigland 
Signed-off-by: Sasha Levin

dlm: make tcp still work in multi-link env

2025-03-18T15:49:22+00:00

This patch bypasses multi-link errors in TCP mode, allowing dlm
to operate on the first tcp link.

Signed-off-by: Heming Zhao 
Signed-off-by: David Teigland

dlm: fix error if active rsb is not hashed

2025-03-01T00:24:21+00:00

If an active rsb is not hashed anymore and this could occur because we
releases and acquired locks we need to signal the followed code that
the lookup failed. Since the lookup was successful, but it isn't part of
the rsb hash anymore we need to signal it by setting error to -EBADR as
dlm_search_rsb_tree() does it.

Cc: stable@vger.kernel.org
Fixes: 5be323b0c64d ("dlm: move dlm_search_rsb_tree() out of lock")
Signed-off-by: Alexander Aring 
Signed-off-by: David Teigland

dlm: fix error if inactive rsb is not hashed

2025-03-01T00:24:21+00:00

If an inactive rsb is not hashed anymore and this could occur because we
releases and acquired locks we need to signal the followed code that the
lookup failed. Since the lookup was successful, but it isn't part of the
rsb hash anymore we need to signal it by setting error to -EBADR as
dlm_search_rsb_tree() does it.

Cc: stable@vger.kernel.org
Fixes: 01fdeca1cc2d ("dlm: use rcu to avoid an extra rsb struct lookup")
Signed-off-by: Alexander Aring 
Signed-off-by: David Teigland

dlm: prevent NPD when writing a positive value to event_done

2025-02-10T19:18:04+00:00

do_uevent returns the value written to event_done. In case it is a
positive value, new_lockspace would undo all the work, and lockspace
would not be set. __dlm_new_lockspace, however, would treat that
positive value as a success due to commit 8511a2728ab8 ("dlm: fix use
count with multiple joins").

Down the line, device_create_lockspace would pass that NULL lockspace to
dlm_find_lockspace_local, leading to a NULL pointer dereference.

Treating such positive values as successes prevents the problem. Given
this has been broken for so long, this is unlikely to break userspace
expectations.

Fixes: 8511a2728ab8 ("dlm: fix use count with multiple joins")
Signed-off-by: Thadeu Lima de Souza Cascardo 
Signed-off-by: David Teigland

dlm: increase max number of links for corosync3/knet

2025-02-10T19:14:05+00:00

This patch increases the maximum number of links that can be
used with corosync3/knet. The majority of the changes are in
user space dlm_tools/dlm_controld.

Signed-off-by: Heming Zhao 
Signed-off-by: David Teigland

dlm: return -ENOENT if no comm was found

2024-12-19T19:11:24+00:00

Currently if no comm can be found dlm_comm_seq() returns -EEXIST which
means entry already exists for a lookup it makes no sense to return
-EEXIST. We change it to -ENOENT. There is no user that will evaluate
the return value on a specific value so this should be fine.

Signed-off-by: Alexander Aring 
Signed-off-by: David Teigland

dlm: fix srcu_read_lock() return type to int

2024-12-19T19:11:23+00:00

The return type of srcu_read_lock() is int and not bool. Whereas we
using the ret variable only to evaluate a bool type of
dlm_lowcomms_con_has_addr() to check if an address is already being set.

Fixes: 6f0b0b5d7ae7 ("fs: dlm: remove dlm_node_addrs lookup list")
Signed-off-by: Alexander Aring 
Signed-off-by: David Teigland

dlm: fix removal of rsb struct that is master and dir record

2024-12-19T19:07:02+00:00

An rsb struct was not being removed in the case where it
was both the master and the dir record.  This case (master
and dir node) was missed in the condition for doing add_scan()
from deactivate_rsb().  Fixing this triggers a related WARN_ON
that needs to be fixed, and requires adjusting where two
del_scan() calls are made.

Fixes: c217adfc8caa ("dlm: fix add_scan and del_scan usage")
Signed-off-by: Alexander Aring 
Signed-off-by: David Teigland

dlm: fix dlm_recover_members refcount on error

2024-11-18T16:05:57+00:00

If dlm_recover_members() fails we don't drop the references of the
previous created root_list that holds and keep all rsbs alive during the
recovery. It might be not an unlikely event because ping_members() could
run into an -EINTR if another recovery progress was triggered again.

Fixes: 3a747f4a2ee8 ("dlm: move rsb root_list to ls_recover() stack")
Signed-off-by: Alexander Aring 
Signed-off-by: David Teigland