Skip to contentSkip to conversation

c/r: Take uts_sem during checkpoint (v2)

PATCH In review Dan Smith 2 participants 4 messages Apr 17, 2009 3 messages not in the archive

lore
Share & tools
lore

Dan Smith

author
#0 lore raw

PATCH 1/1 c/r: Take uts_sem during checkpoint (v2)

Fix the potential for breakage if our UTS changes during checkpoint by grabbing uts_sem and copying those strings to temporary buffers.

Changes in v2:
- Be less stupid about holding the system-wide uts_sem during checkpoint (!)
- Don't hold it during restart
- Hold uts_sem only while copying out the strings
- Calculate the length of the saved buffers outside of the semaphore and avoid the duplicate _len variables by cleaning up the cr_hdr_utsns after the cr_write_string() of the two buffers

---
 checkpoint/ckpt_task.c |   25 ++++++++++++++-----------
 1 files changed, 14 insertions(+), 11 deletions(-)
checkpoint/ckpt_task.c +14 -11
index 4d19e31..30858d2 100644
@@ -171,8 +171,8 @@ static int cr_write_utsns(struct cr_ctx *ctx, struct uts_namespace *uts_ns)
 {
 	struct cr_hdr h;
 	struct cr_hdr_utsns *hh;
-	int domainname_len;
-	int nodename_len;
+	char nodename[__NEW_UTS_LEN + 1];
+	char domainname[__NEW_UTS_LEN + 1];
 	int ret;
 
 	h.type = CR_HDR_UTSNS;
@@ -182,22 +182,25 @@ static int cr_write_utsns(struct cr_ctx *ctx, struct uts_namespace *uts_ns)
 	if (!hh)
 		return -ENOMEM;
 
-	nodename_len = strlen(uts_ns->name.nodename) + 1;
-	domainname_len = strlen(uts_ns->name.domainname) + 1;
+	down_read(&uts_sem);
+	memcpy(nodename, uts_ns->name.nodename, sizeof(nodename));
+	memcpy(domainname, uts_ns->name.domainname, sizeof(nodename));
+	up_read(&uts_sem);
 
-	hh->nodename_len = nodename_len;
-	hh->domainname_len = domainname_len;
+	hh->nodename_len = strlen(nodename) + 1;
+	hh->domainname_len = strlen(domainname) + 1;
 
 	ret = cr_write_obj(ctx, &h, hh);
-	cr_hbuf_put(ctx, sizeof(*hh));
 	if (ret < 0)
-		return ret;
+		goto out;
 
-	ret = cr_write_string(ctx, uts_ns->name.nodename, nodename_len);
+	ret = cr_write_string(ctx, nodename, hh->nodename_len);
 	if (ret < 0)
-		return ret;
+		goto out;
 
-	ret = cr_write_string(ctx, uts_ns->name.domainname, domainname_len);
+	ret = cr_write_string(ctx, domainname, hh->domainname_len);
+ out:
+ 	cr_hbuf_put(ctx, sizeof(*hh));
 	return ret;
 }
 
Signature · 1 line
1.5.6.3

A message that is not in the archive (sent off-list or not yet received): 1239978438-5719-1-git-send-email-danms-r/[email protected]

#2 lore raw
Quoted 5 lines · Dan Smith

Fix the potential for breakage if our UTS changes during checkpoint by grabbing uts_sem and copying those strings to temporary buffers.

Looks good. The only thing I'd add is that you are depending on __NEW_UTS_LEN+1 being something very specific, so if utsname.h gets a change, checkpoint/ckpt_task.c needs a corresponding change. So it would be robust to future code changes if you

#define MAX_UTS_LEN (__NEW_UTS_LEN+1)

in utsname.h, so that anyone expanding the size of hostname doesn't need to look for this usage.

Still,

-serge

Quoted 69 lines

Changes in v2:
- Be less stupid about holding the system-wide uts_sem during checkpoint (!)
- Don't hold it during restart
- Hold uts_sem only while copying out the strings
- Calculate the length of the saved buffers outside of the semaphore and avoid the duplicate _len variables by cleaning up the cr_hdr_utsns after the cr_write_string() of the two buffers

---
 checkpoint/ckpt_task.c |   25 ++++++++++++++-----------
 1 files changed, 14 insertions(+), 11 deletions(-)
checkpoint/ckpt_task.c +14 -11
index 4d19e31..30858d2 100644
@@ -171,8 +171,8 @@ static int cr_write_utsns(struct cr_ctx *ctx, struct uts_namespace *uts_ns)
 {
 	struct cr_hdr h;
 	struct cr_hdr_utsns *hh;
-	int domainname_len;
-	int nodename_len;
+	char nodename[__NEW_UTS_LEN + 1];
+	char domainname[__NEW_UTS_LEN + 1];
 	int ret;
 
 	h.type = CR_HDR_UTSNS;
@@ -182,22 +182,25 @@ static int cr_write_utsns(struct cr_ctx *ctx, struct uts_namespace *uts_ns)
 	if (!hh)
 		return -ENOMEM;
 
-	nodename_len = strlen(uts_ns->name.nodename) + 1;
-	domainname_len = strlen(uts_ns->name.domainname) + 1;
+	down_read(&uts_sem);
+	memcpy(nodename, uts_ns->name.nodename, sizeof(nodename));
+	memcpy(domainname, uts_ns->name.domainname, sizeof(nodename));
+	up_read(&uts_sem);
 
-	hh->nodename_len = nodename_len;
-	hh->domainname_len = domainname_len;
+	hh->nodename_len = strlen(nodename) + 1;
+	hh->domainname_len = strlen(domainname) + 1;
 
 	ret = cr_write_obj(ctx, &h, hh);
-	cr_hbuf_put(ctx, sizeof(*hh));
 	if (ret < 0)
-		return ret;
+		goto out;
 
-	ret = cr_write_string(ctx, uts_ns->name.nodename, nodename_len);
+	ret = cr_write_string(ctx, nodename, hh->nodename_len);
 	if (ret < 0)
-		return ret;
+		goto out;
 
-	ret = cr_write_string(ctx, uts_ns->name.domainname, domainname_len);
+	ret = cr_write_string(ctx, domainname, hh->domainname_len);
+ out:
+ 	cr_hbuf_put(ctx, sizeof(*hh));
 	return ret;
 }
 
Signature · 6 lines
1.5.6.3

_______________________________________________
Containers mailing list
Containers-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org
https://lists.linux-foundation.org/mailman/listinfo/containers

A message that is not in the archive (sent off-list or not yet received): 20090417150529.GB19129-r/[email protected]

author
#4 lore raw

SH> So it would be robust to future code changes if you

SH> #define MAX_UTS_LEN (__NEW_UTS_LEN+1)

SH> in utsname.h, so that anyone expanding the size of hostname SH> doesn't need to look for this usage.

Yeah, I think that came up in a prior round of comments on the UTS stuff, but got dropped when we started depending on strlen() of the names instead. I suppose I should actually do it now... :)

Signature · 3 lines
Dan Smith
IBM Linux Technology Center
email: danms-r/[email protected]

A message that is not in the archive (sent off-list or not yet received): 87ab6fiabr.fsf-FLMGYpZoEPULwtHQx/[email protected]

#6 lore raw
Quoted 10 lines · Dan Smith

SH> So it would be robust to future code changes if you

SH> #define MAX_UTS_LEN (__NEW_UTS_LEN+1)

SH> in utsname.h, so that anyone expanding the size of hostname SH> doesn't need to look for this usage.

Yeah, I think that came up in a prior round of comments on the UTS stuff, but got dropped when we started depending on strlen() of the names instead. I suppose I should actually do it now... :)

Ok, please keep my Ack on the result.

thanks,
-serge