Re: [lustre-devel] [PATCH] staging: lustre: o2iblnd: Fix crash in kiblnd_handle_early_rxs()

[Doug Oucharek <doucharek@cray.com>]

[-- Attachment #1: Type: text/plain, Size: 2529 bytes --]

I did a v2 of this patch already.

Changing to the safe version of the list macros is a mixed bag.

Doug

On May 10, 2018, at 5:53 PM, NeilBrown <neilb@suse.com<mailto:neilb@suse.com>> wrote:

On Wed, May 09 2018, Doug Oucharek wrote:

Under upstream staging commit 5a2ca43fa54f561c252c2, the list handling
code in kiblnd_handle_early_rxs() got changed to list_for_each_safe().
That protects against the current thread from deleting the current entry
it is looking at. It does not protect against another thread from deleting
the next item in the list (which the tmp variable points to). The way this
routine holds then releases a lock opens the door to other threads doing
just that.

This patch reverts this commit on this routine.

Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-9886
Signed-off-by: Doug Oucharek <dougso@me.com<mailto:dougso@me.com>>
---
drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd_cb.c | 5 +++--
1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd_cb.c b/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd_cb.c
index 32fa8ca..6148fbb 100644
--- a/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd_cb.c
+++ b/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd_cb.c
@@ -1965,13 +1965,14 @@ static int kiblnd_resolve_addr(struct rdma_cm_id *cmid,
{
unsigned long flags;
struct kib_rx *rx;
- struct kib_rx *tmp;

LASSERT(!in_interrupt());
LASSERT(conn->ibc_state >= IBLND_CONN_ESTABLISHED);

write_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
- list_for_each_entry_safe(rx, tmp, &conn->ibc_early_rxs, rx_list) {
+ while (!list_empty(&conn->ibc_early_rxs)) {
+ rx = list_entry(conn->ibc_early_rxs.next,
+ kib_rx_t, rx_list);
Should be:
                        struct kib_tx

Otherwise,
Reviewed-by: NeilBrown <neilb@suse.com<mailto:neilb@suse.com>>

Those "convert lots of list_for_each" things really do need
careful review, don't they :-(

Thanks,
NeilBrown

list_del(&rx->rx_list);
write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);

--
1.8.3.1

_______________________________________________
lustre-devel mailing list
lustre-devel@lists.lustre.org<mailto:lustre-devel@lists.lustre.org>
http://lists.lustre.org/listinfo.cgi/lustre-devel-lustre.org
_______________________________________________
lustre-devel mailing list
lustre-devel@lists.lustre.org<mailto:lustre-devel@lists.lustre.org>
http://lists.lustre.org/listinfo.cgi/lustre-devel-lustre.org


[-- Attachment #2: Type: text/html, Size: 16027 bytes --]