diff mbox series

[net,2/5] net/smc: fix incorrect barrier usage

Message ID 1697009600-22367-3-git-send-email-alibuda@linux.alibaba.com (mailing list archive)
State Changes Requested
Delegated to: Netdev Maintainers
Headers show
Series net/smc: bugfixs for smc-r | expand

Checks

Context Check Description
netdev/series_format success Posting correctly formatted
netdev/tree_selection success Clearly marked for net
netdev/fixes_present success Fixes tag present in non-next series
netdev/header_inline success No static functions without inline keyword in header files
netdev/build_32bit success Errors and warnings before: 1363 this patch: 1363
netdev/cc_maintainers warning 4 maintainers not CCed: tonylu@linux.alibaba.com pabeni@redhat.com edumazet@google.com guwen@linux.alibaba.com
netdev/build_clang success Errors and warnings before: 1387 this patch: 1387
netdev/verify_signedoff success Signed-off-by tag matches author and committer
netdev/deprecated_api success None detected
netdev/check_selftest success No net selftest shell script
netdev/verify_fixes success Fixes tag looks correct
netdev/build_allmodconfig_warn success Errors and warnings before: 1387 this patch: 1387
netdev/checkpatch warning WARNING: line length of 82 exceeds 80 columns WARNING: line length of 87 exceeds 80 columns
netdev/kdoc success Errors and warnings before: 0 this patch: 0
netdev/source_inline success Was 0 now: 0

Commit Message

D. Wythe Oct. 11, 2023, 7:33 a.m. UTC 
From: "D. Wythe" <alibuda@linux.alibaba.com>

This patch add explicit CPU barrier to ensure memory
consistency rather than compiler barrier.

Besides, the atomicity between READ_ONCE and cmpxhcg cannot
be guaranteed, so we need to use atomic ops. The simple way
is to replace READ_ONCE with xchg.

Fixes: 475f9ff63ee8 ("net/smc: fix application data exception")
Co-developed-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: D. Wythe <alibuda@linux.alibaba.com>
Links: https://lore.kernel.org/netdev/1b7c95be-d3d9-53c3-3152-cd835314d37c@linux.ibm.com/T/
---
 net/smc/smc_core.c | 21 +++++++++++++--------
 1 file changed, 13 insertions(+), 8 deletions(-)

Comments

Heiko Carstens Oct. 11, 2023, 8:44 a.m. UTC | #1 
On Wed, Oct 11, 2023 at 03:33:17PM +0800, D. Wythe wrote:
> From: "D. Wythe" <alibuda@linux.alibaba.com>
> 
> This patch add explicit CPU barrier to ensure memory
> consistency rather than compiler barrier.
> 
> Besides, the atomicity between READ_ONCE and cmpxhcg cannot
> be guaranteed, so we need to use atomic ops. The simple way
> is to replace READ_ONCE with xchg.
> 
> Fixes: 475f9ff63ee8 ("net/smc: fix application data exception")
> Co-developed-by: Heiko Carstens <hca@linux.ibm.com>
> Signed-off-by: Heiko Carstens <hca@linux.ibm.com>

^^^
I did not Co-develop this, nor did I provide an explicit Signed-off-by.
Please don't add Signed-off-by statements which have not been explicitly
agreed on.
D. Wythe Oct. 11, 2023, 8:57 a.m. UTC | #2 
On 10/11/23 4:44 PM, Heiko Carstens wrote:
> I did not Co-develop this, nor did I provide an explicit Signed-off-by.
> Please don't add Signed-off-by statements which have not been explicitly
> agreed on.

Sorry for that, I have used the wrong tag, Reported by might be more 
appropriate .
I will remove this in the next version, sorry to bother you again.

Best wishes,
D. Wythe
diff mbox series

Patch

diff --git a/net/smc/smc_core.c b/net/smc/smc_core.c
index d520ee6..cc7d72e 100644
--- a/net/smc/smc_core.c
+++ b/net/smc/smc_core.c
@@ -1133,9 +1133,10 @@  static void smcr_buf_unuse(struct smc_buf_desc *buf_desc, bool is_rmb,
 
 		smc_buf_free(lgr, is_rmb, buf_desc);
 	} else {
-		/* memzero_explicit provides potential memory barrier semantics */
-		memzero_explicit(buf_desc->cpu_addr, buf_desc->len);
-		WRITE_ONCE(buf_desc->used, 0);
+		memset(buf_desc->cpu_addr, 0, buf_desc->len);
+		/* make sure buf_desc->used not be reordered ahead */
+		smp_mb__before_atomic();
+		xchg(&buf_desc->used, 0);
 	}
 }
 
@@ -1146,17 +1147,21 @@  static void smc_buf_unuse(struct smc_connection *conn,
 		if (!lgr->is_smcd && conn->sndbuf_desc->is_vm) {
 			smcr_buf_unuse(conn->sndbuf_desc, false, lgr);
 		} else {
-			memzero_explicit(conn->sndbuf_desc->cpu_addr, conn->sndbuf_desc->len);
-			WRITE_ONCE(conn->sndbuf_desc->used, 0);
+			memset(conn->sndbuf_desc->cpu_addr, 0, conn->sndbuf_desc->len);
+			/* make sure buf_desc->used not be reordered ahead */
+			smp_mb__before_atomic();
+			xchg(&conn->sndbuf_desc->used, 0);
 		}
 	}
 	if (conn->rmb_desc) {
 		if (!lgr->is_smcd) {
 			smcr_buf_unuse(conn->rmb_desc, true, lgr);
 		} else {
-			memzero_explicit(conn->rmb_desc->cpu_addr,
-					 conn->rmb_desc->len + sizeof(struct smcd_cdc_msg));
-			WRITE_ONCE(conn->rmb_desc->used, 0);
+			memset(conn->rmb_desc->cpu_addr, 0,
+			       conn->rmb_desc->len + sizeof(struct smcd_cdc_msg));
+			/* make sure buf_desc->used not be reordered ahead */
+			smp_mb__before_atomic();
+			xchg(&conn->rmb_desc->used, 0);
 		}
 	}
 }