The BUG_ON() recently introduced in lpfc_sli_ringtxcmpl_put()
is hit in the lpfc_els_abort() > lpfc_sli_issue_abort_iotag()
> lpfc_sli_abort_iotag_issue() function path [similar names],
due to 'piocb->vport == NULL':
BUG_ON(!piocb || !piocb->vport);
This happens because lpfc_sli_abort_iotag_issue() doesn't set
the 'abtsiocbp->vport' pointer -- but this is not the problem.
Previously, lpfc_sli_ringtxcmpl_put() accessed 'piocb->vport'
only if 'piocb->iocb.ulpCommand' is neither CMD_ABORT_XRI_CN
nor CMD_CLOSE_XRI_CN, which are the only possible values for
lpfc_sli_abort_iotag_issue():
lpfc_sli_ringtxcmpl_put():
if ((unlikely(pring->ringno == LPFC_ELS_RING)) &&
(piocb->iocb.ulpCommand != CMD_ABORT_XRI_CN) &&
(piocb->iocb.ulpCommand != CMD_CLOSE_XRI_CN) &&
(!(piocb->vport->load_flag & FC_UNLOADING)))
lpfc_sli_abort_iotag_issue():
if (phba->link_state >= LPFC_LINK_UP)
iabt->ulpCommand = CMD_ABORT_XRI_CN;
else
iabt->ulpCommand = CMD_CLOSE_XRI_CN;
So, this function path would not have hit this possible NULL
pointer dereference before.
In order to fix this regression, move the second part of the
BUG_ON() check prior to the pointer dereference that it does
check for.
For reference, this is the stack trace observed. The problem
happened because an unsolicited event was received - a PLOGI
was received after our PLOGI was issued but not yet complete,
so the discovery state machine goes on to sw-abort our PLOGI.
kernel BUG at drivers/scsi/lpfc/lpfc_sli.c:1326!
Oops: Exception in kernel mode, sig: 5 [#1]
<...>
NIP [...] lpfc_sli_ringtxcmpl_put+0x1c/0xf0 [lpfc]
LR [...] __lpfc_sli_issue_iocb_s4+0x188/0x200 [lpfc]
Call Trace:
[...] [...] __lpfc_sli_issue_iocb_s4+0xb0/0x200 [lpfc] (unreliable)
[...] [...] lpfc_sli_issue_abort_iotag+0x2b4/0x350 [lpfc]
[...] [...] lpfc_els_abort+0x1a8/0x4a0 [lpfc]
[...] [...] lpfc_rcv_plogi+0x6d4/0x700 [lpfc]
[...] [...] lpfc_rcv_plogi_plogi_issue+0xd8/0x1d0 [lpfc]
[...] [...] lpfc_disc_state_machine+0xc0/0x2b0 [lpfc]
[...] [...] lpfc_els_unsol_buffer+0xcc0/0x26c0 [lpfc]
[...] [...] lpfc_els_unsol_event+0xa8/0x220 [lpfc]
[...] [...] lpfc_complete_unsol_iocb+0xb8/0x138 [lpfc]
[...] [...] lpfc_sli4_handle_received_buffer+0x6a0/0xec0 [lpfc]
[...] [...] lpfc_sli_handle_slow_ring_event_s4+0x1c4/0x240 [lpfc]
[...] [...] lpfc_sli_handle_slow_ring_event+0x24/0x40 [lpfc]
[...] [...] lpfc_do_work+0xd88/0x1970 [lpfc]
[...] [...] kthread+0x108/0x130
[...] [...] ret_from_kernel_thread+0x5c/0xbc
<...>
Cc: [email protected] # v4.8
Fixes: 22466da5b4b7 ("lpfc: Fix possible NULL pointer dereference")
Signed-off-by: Mauricio Faria de Oliveira <[email protected]>
---
drivers/scsi/lpfc/lpfc_sli.c | 14 ++++++++------
1 file changed, 8 insertions(+), 6 deletions(-)
diff --git a/drivers/scsi/lpfc/lpfc_sli.c b/drivers/scsi/lpfc/lpfc_sli.c
index c5326055beee..f4f77c5b0c83 100644
--- a/drivers/scsi/lpfc/lpfc_sli.c
+++ b/drivers/scsi/lpfc/lpfc_sli.c
@@ -1323,18 +1323,20 @@ struct lpfc_iocbq *
{
lockdep_assert_held(&phba->hbalock);
- BUG_ON(!piocb || !piocb->vport);
+ BUG_ON(!piocb);
list_add_tail(&piocb->list, &pring->txcmplq);
piocb->iocb_flag |= LPFC_IO_ON_TXCMPLQ;
if ((unlikely(pring->ringno == LPFC_ELS_RING)) &&
(piocb->iocb.ulpCommand != CMD_ABORT_XRI_CN) &&
- (piocb->iocb.ulpCommand != CMD_CLOSE_XRI_CN) &&
- (!(piocb->vport->load_flag & FC_UNLOADING)))
- mod_timer(&piocb->vport->els_tmofunc,
- jiffies +
- msecs_to_jiffies(1000 * (phba->fc_ratov << 1)));
+ (piocb->iocb.ulpCommand != CMD_CLOSE_XRI_CN)) {
+ BUG_ON(!piocb->vport);
+ if (!(piocb->vport->load_flag & FC_UNLOADING))
+ mod_timer(&piocb->vport->els_tmofunc,
+ jiffies +
+ msecs_to_jiffies(1000 * (phba->fc_ratov << 1)));
+ }
return 0;
}
--
1.8.3.1
Due credit; an oversight.
On 11/23/2016 10:33 AM, Mauricio Faria de Oliveira wrote:
Reported-by: Harsha Thyagaraja <[email protected]>
> Cc: [email protected] # v4.8
> Fixes: 22466da5b4b7 ("lpfc: Fix possible NULL pointer dereference")
> Signed-off-by: Mauricio Faria de Oliveira <[email protected]>
--
Mauricio Faria de Oliveira
IBM Linux Technology Center
On Wed, Nov 23, 2016 at 10:33:19AM -0200, Mauricio Faria de Oliveira wrote:
> The BUG_ON() recently introduced in lpfc_sli_ringtxcmpl_put()
> is hit in the lpfc_els_abort() > lpfc_sli_issue_abort_iotag()
> > lpfc_sli_abort_iotag_issue() function path [similar names],
> due to 'piocb->vport == NULL':
>
> BUG_ON(!piocb || !piocb->vport);
>
> This happens because lpfc_sli_abort_iotag_issue() doesn't set
> the 'abtsiocbp->vport' pointer -- but this is not the problem.
>
> Previously, lpfc_sli_ringtxcmpl_put() accessed 'piocb->vport'
> only if 'piocb->iocb.ulpCommand' is neither CMD_ABORT_XRI_CN
> nor CMD_CLOSE_XRI_CN, which are the only possible values for
> lpfc_sli_abort_iotag_issue():
>
> lpfc_sli_ringtxcmpl_put():
>
> if ((unlikely(pring->ringno == LPFC_ELS_RING)) &&
> (piocb->iocb.ulpCommand != CMD_ABORT_XRI_CN) &&
> (piocb->iocb.ulpCommand != CMD_CLOSE_XRI_CN) &&
> (!(piocb->vport->load_flag & FC_UNLOADING)))
>
> lpfc_sli_abort_iotag_issue():
>
> if (phba->link_state >= LPFC_LINK_UP)
> iabt->ulpCommand = CMD_ABORT_XRI_CN;
> else
> iabt->ulpCommand = CMD_CLOSE_XRI_CN;
>
> So, this function path would not have hit this possible NULL
> pointer dereference before.
>
> In order to fix this regression, move the second part of the
> BUG_ON() check prior to the pointer dereference that it does
> check for.
>
> For reference, this is the stack trace observed. The problem
> happened because an unsolicited event was received - a PLOGI
> was received after our PLOGI was issued but not yet complete,
> so the discovery state machine goes on to sw-abort our PLOGI.
>
> kernel BUG at drivers/scsi/lpfc/lpfc_sli.c:1326!
> Oops: Exception in kernel mode, sig: 5 [#1]
> <...>
> NIP [...] lpfc_sli_ringtxcmpl_put+0x1c/0xf0 [lpfc]
> LR [...] __lpfc_sli_issue_iocb_s4+0x188/0x200 [lpfc]
> Call Trace:
> [...] [...] __lpfc_sli_issue_iocb_s4+0xb0/0x200 [lpfc] (unreliable)
> [...] [...] lpfc_sli_issue_abort_iotag+0x2b4/0x350 [lpfc]
> [...] [...] lpfc_els_abort+0x1a8/0x4a0 [lpfc]
> [...] [...] lpfc_rcv_plogi+0x6d4/0x700 [lpfc]
> [...] [...] lpfc_rcv_plogi_plogi_issue+0xd8/0x1d0 [lpfc]
> [...] [...] lpfc_disc_state_machine+0xc0/0x2b0 [lpfc]
> [...] [...] lpfc_els_unsol_buffer+0xcc0/0x26c0 [lpfc]
> [...] [...] lpfc_els_unsol_event+0xa8/0x220 [lpfc]
> [...] [...] lpfc_complete_unsol_iocb+0xb8/0x138 [lpfc]
> [...] [...] lpfc_sli4_handle_received_buffer+0x6a0/0xec0 [lpfc]
> [...] [...] lpfc_sli_handle_slow_ring_event_s4+0x1c4/0x240 [lpfc]
> [...] [...] lpfc_sli_handle_slow_ring_event+0x24/0x40 [lpfc]
> [...] [...] lpfc_do_work+0xd88/0x1970 [lpfc]
> [...] [...] kthread+0x108/0x130
> [...] [...] ret_from_kernel_thread+0x5c/0xbc
> <...>
>
> Cc: [email protected] # v4.8
> Fixes: 22466da5b4b7 ("lpfc: Fix possible NULL pointer dereference")
> Signed-off-by: Mauricio Faria de Oliveira <[email protected]>
> ---
Looks good and sorry for the bug,
Reviewed-by: Johannes Thumshirn <[email protected]>
--
Johannes Thumshirn Storage
[email protected] +49 911 74053 689
SUSE LINUX GmbH, Maxfeldstr. 5, 90409 N?rnberg
GF: Felix Imend?rffer, Jane Smithard, Graham Norton
HRB 21284 (AG N?rnberg)
Key fingerprint = EC38 9CAB C2C4 F25D 8600 D0D0 0393 969D 2D76 0850
On 11/23/2016 12:12 PM, Johannes Thumshirn wrote:
> Looks good and sorry for the bug,
> Reviewed-by: Johannes Thumshirn <[email protected]>
Thanks for the quick review. Not a problem!
This problem turned out to be a good learning exercise. :)
--
Mauricio Faria de Oliveira
IBM Linux Technology Center
>>>>> "Mauricio" == Mauricio Faria de Oliveira <[email protected]> writes:
Mauricio> The BUG_ON() recently introduced in lpfc_sli_ringtxcmpl_put()
Mauricio> is hit in the lpfc_els_abort() > lpfc_sli_issue_abort_iotag()
Mauricio> > lpfc_sli_abort_iotag_issue() function path [similar names],
Mauricio> due to 'piocb->vport == NULL':
Applied to 4.9/scsi-fixes.
--
Martin K. Petersen Oracle Linux Engineering