On 31.08.22 17:53, liuyacan@xxxxxxxxxxxxxxxx wrote:
From: Yacan Liu <liuyacan@xxxxxxxxxxxxxxxx>
After modifying the QP to the Error state, all RX WR would be completed
with WC in IB_WC_WR_FLUSH_ERR status. Current implementation does not
wait for it is done, but destroy the QP and free the link group directly.
So there is a risk that accessing the freed memory in tasklet context.
Here is a crash example:
BUG: unable to handle page fault for address: ffffffff8f220860
#PF: supervisor write access in kernel mode
#PF: error_code(0x0002) - not-present page
PGD f7300e067 P4D f7300e067 PUD f7300f063 PMD 8c4e45063 PTE 800ffff08c9df060
Oops: 0002 [#1] SMP PTI
CPU: 1 PID: 0 Comm: swapper/1 Kdump: loaded Tainted: G S OE 5.10.0-0607+ #23
Hardware name: Inspur NF5280M4/YZMB-00689-101, BIOS 4.1.20 07/09/2018
RIP: 0010:native_queued_spin_lock_slowpath+0x176/0x1b0
Code: f3 90 48 8b 32 48 85 f6 74 f6 eb d5 c1 ee 12 83 e0 03 83 ee 01 48 c1 e0 05 48 63 f6 48 05 00 c8 02 00 48 03 04 f5 00 09 98 8e <48> 89 10 8b 42 08 85 c0 75 09 f3 90 8b 42 08 85 c0 74 f7 48 8b 32
RSP: 0018:ffffb3b6c001ebd8 EFLAGS: 00010086
RAX: ffffffff8f220860 RBX: 0000000000000246 RCX: 0000000000080000
RDX: ffff91db1f86c800 RSI: 000000000000173c RDI: ffff91db62bace00
RBP: ffff91db62bacc00 R08: 0000000000000000 R09: c00000010000028b
R10: 0000000000055198 R11: ffffb3b6c001ea58 R12: ffff91db80e05010
R13: 000000000000000a R14: 0000000000000006 R15: 0000000000000040
FS: 0000000000000000(0000) GS:ffff91db1f840000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: ffffffff8f220860 CR3: 00000001f9580004 CR4: 00000000003706e0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
Call Trace:
<IRQ>
_raw_spin_lock_irqsave+0x30/0x40
mlx5_ib_poll_cq+0x4c/0xc50 [mlx5_ib]
smc_wr_rx_tasklet_fn+0x56/0xa0 [smc]
tasklet_action_common.isra.21+0x66/0x100
__do_softirq+0xd5/0x29c
asm_call_irq_on_stack+0x12/0x20
</IRQ>
do_softirq_own_stack+0x37/0x40
irq_exit_rcu+0x9d/0xa0
sysvec_call_function_single+0x34/0x80
asm_sysvec_call_function_single+0x12/0x20
Fixes: bd4ad57718cc ("smc: initialize IB transport incl. PD, MR, QP, CQ, event, WR")
Signed-off-by: Yacan Liu <liuyacan@xxxxxxxxxxxxxxxx>
---
Chagen in v4:
-- Remove the rx_drain flag because smc_wr_rx_post() may not have been called.
-- Remove timeout.
Change in v3:
-- Tune commit message (Signed-Off tag, Fixes tag).
Tune code to avoid column length exceeding.
Change in v2:
-- Fix some compile warnings and errors.
---
net/smc/smc_core.c | 2 ++
net/smc/smc_core.h | 2 ++
net/smc/smc_wr.c | 9 +++++++++
net/smc/smc_wr.h | 1 +
4 files changed, 14 insertions(+)
diff --git a/net/smc/smc_core.c b/net/smc/smc_core.c
index ff49a11f5..f92a916e9 100644
--- a/net/smc/smc_core.c
+++ b/net/smc/smc_core.c
@@ -757,6 +757,7 @@ int smcr_link_init(struct smc_link_group *lgr, struct smc_link *lnk,
lnk->lgr = lgr;
smc_lgr_hold(lgr); /* lgr_put in smcr_link_clear() */
lnk->link_idx = link_idx;
+ lnk->wr_rx_id_compl = 0;
smc_ibdev_cnt_inc(lnk);
smcr_copy_dev_info_to_link(lnk);
atomic_set(&lnk->conn_cnt, 0);
@@ -1269,6 +1270,7 @@ void smcr_link_clear(struct smc_link *lnk, bool log)
smcr_buf_unmap_lgr(lnk);
smcr_rtoken_clear_link(lnk);
smc_ib_modify_qp_error(lnk);
+ smc_wr_drain_cq(lnk);
smc_wr_free_link(lnk);
smc_ib_destroy_queue_pair(lnk);
smc_ib_dealloc_protection_domain(lnk);
diff --git a/net/smc/smc_core.h b/net/smc/smc_core.h
index fe8b524ad..285f9bd8e 100644
--- a/net/smc/smc_core.h
+++ b/net/smc/smc_core.h
@@ -115,8 +115,10 @@ struct smc_link {
dma_addr_t wr_rx_dma_addr; /* DMA address of wr_rx_bufs */
dma_addr_t wr_rx_v2_dma_addr; /* DMA address of v2 rx buf*/
u64 wr_rx_id; /* seq # of last recv WR */
+ u64 wr_rx_id_compl; /* seq # of last completed WR */
u32 wr_rx_cnt; /* number of WR recv buffers */
unsigned long wr_rx_tstamp; /* jiffies when last buf rx */
+ wait_queue_head_t wr_rx_empty_wait; /* wait for RQ empty */
struct ib_reg_wr wr_reg; /* WR register memory region */
wait_queue_head_t wr_reg_wait; /* wait for wr_reg result */
diff --git a/net/smc/smc_wr.c b/net/smc/smc_wr.c
index 26f8f240d..bc8793803 100644
--- a/net/smc/smc_wr.c
+++ b/net/smc/smc_wr.c
@@ -454,6 +454,7 @@ static inline void smc_wr_rx_process_cqes(struct ib_wc wc[], int num)
for (i = 0; i < num; i++) {
link = wc[i].qp->qp_context;
+ link->wr_rx_id_compl = wc[i].wr_id;
if (wc[i].status == IB_WC_SUCCESS) {
link->wr_rx_tstamp = jiffies;
smc_wr_rx_demultiplex(&wc[i]);
@@ -465,6 +466,8 @@ static inline void smc_wr_rx_process_cqes(struct ib_wc wc[], int num)
case IB_WC_RNR_RETRY_EXC_ERR:
case IB_WC_WR_FLUSH_ERR:
smcr_link_down_cond_sched(link);
+ if (link->wr_rx_id_compl == link->wr_rx_id)
+ wake_up(&link->wr_rx_empty_wait);
break;
default:
smc_wr_rx_post(link); /* refill WR RX */
@@ -631,6 +634,11 @@ static void smc_wr_init_sge(struct smc_link *lnk)
lnk->wr_reg.access = IB_ACCESS_LOCAL_WRITE | IB_ACCESS_REMOTE_WRITE;
}
+void smc_wr_drain_cq(struct smc_link *lnk)
+{
+ wait_event(lnk->wr_rx_empty_wait, lnk->wr_rx_id_compl == lnk->wr_rx_id);
+}
+
void smc_wr_free_link(struct smc_link *lnk)
{
struct ib_device *ibdev;
@@ -889,6 +897,7 @@ int smc_wr_create_link(struct smc_link *lnk)
atomic_set(&lnk->wr_tx_refcnt, 0);
init_waitqueue_head(&lnk->wr_reg_wait);
atomic_set(&lnk->wr_reg_refcnt, 0);
+ init_waitqueue_head(&lnk->wr_rx_empty_wait);
return rc;
dma_unmap:
diff --git a/net/smc/smc_wr.h b/net/smc/smc_wr.h
index a54e90a11..5ca5086ae 100644
--- a/net/smc/smc_wr.h
+++ b/net/smc/smc_wr.h
@@ -101,6 +101,7 @@ static inline int smc_wr_rx_post(struct smc_link *link)
int smc_wr_create_link(struct smc_link *lnk);
int smc_wr_alloc_link_mem(struct smc_link *lnk);
int smc_wr_alloc_lgr_mem(struct smc_link_group *lgr);
+void smc_wr_drain_cq(struct smc_link *lnk);
void smc_wr_free_link(struct smc_link *lnk);
void smc_wr_free_link_mem(struct smc_link *lnk);
void smc_wr_free_lgr_mem(struct smc_link_group *lgr);
Thank you @Yacan for the effort to improve our code! And Thank you @Tony
for such valuable suggestions and testing!
I like the modification of this version. However, this is not a fix
patch to upstream, since the patches "[PATCH net-next v2 00/10] optimize
the parallelism of SMC-R connections" are still not applied. My
sugguestions:
- Please talk to the author (D. Wythe <alibuda@xxxxxxxxxxxxxxxxx>) of
those patches I mentioned above, and ask if he can take your patch as a
part of the patch serie
- Fix patches should go to net-next
- Please send always send your new version separately, rather than as
reply to your previous version. That makes people confused.
