Hi Christian,
Could you pick this up, please?
David
---
afs: Fix fileserver rotation getting stuck
Fix the fileserver rotation code in a couple of ways:
(1) op->server_states is an array, not a pointer to a single record, so
fix the places that access it to index it.
(2) In the places that go through an address list to work out which one
has the best priority, fix the loops to skip known failed addresses.
Without this, the rotation algorithm may get stuck on addresses that are
inaccessible or don't respond.
This can be triggered manually by finding a server that advertises a
non-routable address and giving it a higher priority, eg.:
echo "add udp 192.168.0.0/16 3000" >/proc/fs/afs/addr_prefs
if the server, say, includes the address 192.168.7.7 in its address list,
and then attempting to access a volume on that server.
Fixes: 495f2ae9e355 ("afs: Fix fileserver rotation")
Signed-off-by: David Howells <[email protected]>
cc: Marc Dionne <[email protected]>
cc: [email protected]
Link: https://lore.kernel.org/r/[email protected]/ # v1
---
Changes
=======
ver #2)
- Use the untried address set precomputed in the 'set' variable.
fs/afs/rotate.c | 8 ++++++--
1 file changed, 6 insertions(+), 2 deletions(-)
diff --git a/fs/afs/rotate.c b/fs/afs/rotate.c
index ed04bd1eeae8..ed09d4d4c211 100644
--- a/fs/afs/rotate.c
+++ b/fs/afs/rotate.c
@@ -541,11 +541,13 @@ bool afs_select_fileserver(struct afs_operation *op)
test_bit(AFS_SE_EXCLUDED, &se->flags) ||
!test_bit(AFS_SERVER_FL_RESPONDING, &s->flags))
continue;
- es = op->server_states->endpoint_state;
+ es = op->server_states[i].endpoint_state;
sal = es->addresses;
afs_get_address_preferences_rcu(op->net, sal);
for (j = 0; j < sal->nr_addrs; j++) {
+ if (es->failed_set & (1 << j))
+ continue;
if (!sal->addrs[j].peer)
continue;
if (sal->addrs[j].prio > best_prio) {
@@ -605,6 +607,8 @@ bool afs_select_fileserver(struct afs_operation *op)
best_prio = -1;
addr_index = 0;
for (i = 0; i < alist->nr_addrs; i++) {
+ if (!(set & (1 << i)))
+ continue;
if (alist->addrs[i].prio > best_prio) {
addr_index = i;
best_prio = alist->addrs[i].prio;
@@ -674,7 +678,7 @@ bool afs_select_fileserver(struct afs_operation *op)
for (i = 0; i < op->server_list->nr_servers; i++) {
struct afs_endpoint_state *estate;
- estate = op->server_states->endpoint_state;
+ estate = op->server_states[i].endpoint_state;
error = READ_ONCE(estate->error);
if (error < 0)
afs_op_accumulate_error(op, error, estate->abort_code);
On Fri, 03 May 2024 15:22:32 +0100, David Howells wrote:
> Could you pick this up, please?
>
> David
>
>
Applied to the vfs.fixes branch of the vfs/vfs.git tree.
Patches in the vfs.fixes branch should appear in linux-next soon.
Please report any outstanding bugs that were missed during review in a
new review to the original patch series allowing us to drop it.
It's encouraged to provide Acked-bys and Reviewed-bys even though the
patch has now been applied. If possible patch trailers will be updated.
Note that commit hashes shown below are subject to change due to rebase,
trailer updates or similar. If in doubt, please check the listed branch.
tree: https://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs.git
branch: vfs.fixes
[1/1] afs: Fix fileserver rotation getting stuck
https://git.kernel.org/vfs/vfs/c/21965b764cec