c/r: Take uts_sem during checkpoint and restart
PATCH 1/1 c/r: Take uts_sem during checkpoint and restart
Fix the potential for breakage if our UTS changes during checkpoint or restart by grabbing uts_sem during these operations.
--- checkpoint/ckpt_task.c | 12 ++++++++---- checkpoint/rstr_task.c | 17 +++++++++++------ 2 files changed, 19 insertions(+), 10 deletions(-)
checkpoint/ckpt_task.c +8 -4
@@ -173,14 +173,16 @@ static int cr_write_utsns(struct cr_ctx *ctx, struct uts_namespace *uts_ns)struct cr_hdr_utsns *hh;int domainname_len;int nodename_len; -int ret; +int ret = -ENOMEM ; + +down_read (&uts_sem); h.type =CR_HDR_UTSNS ; h.len =sizeof (*hh); hh =cr_hbuf_get (ctx,sizeof (*hh));if (!hh) -return -ENOMEM ; + goto out; nodename_len =strlen (uts_ns->name.nodename) +1 ; domainname_len =strlen (uts_ns->name.domainname) +1 ;
@@ -191,13 +193,15 @@ static int cr_write_utsns(struct cr_ctx *ctx, struct uts_namespace *uts_ns) 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);if (ret <0 ) -return ret; + goto out; ret =cr_write_string (ctx, uts_ns->name.domainname, domainname_len); + out: +up_read (&uts_sem);return ret; }
checkpoint/rstr_task.c +11 -6
@@ -171,32 +171,37 @@ static int cr_read_utsns(struct cr_ctx *ctx) {struct cr_hdr_utsns *hh;struct uts_namespace *ns; -int ret; +int ret = -ENOMEM ; + +down_write (&uts_sem); hh =cr_hbuf_get (ctx,sizeof (*hh));if (!hh) -return -ENOMEM ; + goto out; ret =cr_read_obj_type (ctx, hh,sizeof (*hh),CR_HDR_UTSNS );if (ret <0 ) - goto out; + goto out_hbuf; ret = -EINVAL ;if (hh->nodename_len >sizeof (ns->name.nodename) || hh->domainname_len >sizeof (ns->name.domainname)) - goto out; + goto out_hbuf; ns = current->nsproxy->uts_ns;memset (ns->name.nodename,0 ,sizeof (ns->name.nodename)); ret =cr_read_string (ctx, ns->name.nodename, hh->nodename_len);if (ret <0 ) - goto out; + goto out_hbuf;memset (ns->name.domainname,0 ,sizeof (ns->name.domainname)); ret =cr_read_string (ctx, ns->name.domainname, hh->domainname_len); + + out_hbuf: +cr_hbuf_put (ctx,sizeof (*hh)); out: -cr_hbuf_put (ctx,sizeof (*hh)); +up_write (&uts_sem);return ret; }
Signature · 1 line
1.5.6.3
A message that is not in the archive (sent off-list or not yet received): 1239897585-19246-1-git-send-email-danms-r/[email protected]
Quoted 2 lines · Dan Smith
Fix the potential for breakage if our UTS changes during checkpoint or restart by grabbing
uts_semduring these operations.
Thanks, but I think you're waving too big a stick here.
At checkpoint, you only need the uts_sem right around the reading (and length-checking) of uts_ns->name. You're keeping it around writing into the checkpoint image (and cr_hbuf_get).
At restart, you shouldn't actually need uts_sem - the nsproxy is not yet available to anyone else, so there is no chance of a race, right? Hmm, well maybe I'm being overly optimistic, since restart is coordinated by userspace which could mess up and somehow run sethostname(2) while it's supposed to be running sys_restart(). But then sys_restart() wouldn't yet have hooked up the task to the new nsproxy.
So yeah, I don't think you need uts_sem at restart.
thanks,
-serge
Quoted 103 lines
Signed-off-by Dan Smith <danms-r/[email protected]>--- checkpoint/ckpt_task.c | 12 ++++++++---- checkpoint/rstr_task.c | 17 +++++++++++------ 2 files changed, 19 insertions(+), 10 deletions(-)checkpoint/ckpt_task.c +8
-4@@ -173,14 +173,16 @@ static int cr_write_utsns(struct cr_ctx *ctx, struct uts_namespace *uts_ns)struct cr_hdr_utsns *hh;int domainname_len;int nodename_len; -int ret; +int ret = -ENOMEM ; + +down_read (&uts_sem); h.type =CR_HDR_UTSNS ; h.len =sizeof (*hh); hh =cr_hbuf_get (ctx,sizeof (*hh));if (!hh) -return -ENOMEM ; + goto out; nodename_len =strlen (uts_ns->name.nodename) +1 ; domainname_len =strlen (uts_ns->name.domainname) +1 ;@@ -191,13 +193,15 @@ static int cr_write_utsns(struct cr_ctx *ctx, struct uts_namespace *uts_ns) 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);if (ret <0 ) -return ret; + goto out; ret =cr_write_string (ctx, uts_ns->name.domainname, domainname_len); + out: +up_read (&uts_sem);return ret; }checkpoint/rstr_task.c +11
-6@@ -171,32 +171,37 @@ static int cr_read_utsns(struct cr_ctx *ctx) {struct cr_hdr_utsns *hh;struct uts_namespace *ns; -int ret; +int ret = -ENOMEM ; + +down_write (&uts_sem); hh =cr_hbuf_get (ctx,sizeof (*hh));if (!hh) -return -ENOMEM ; + goto out; ret =cr_read_obj_type (ctx, hh,sizeof (*hh),CR_HDR_UTSNS );if (ret <0 ) - goto out; + goto out_hbuf; ret = -EINVAL ;if (hh->nodename_len >sizeof (ns->name.nodename) || hh->domainname_len >sizeof (ns->name.domainname)) - goto out; + goto out_hbuf; ns = current->nsproxy->uts_ns;memset (ns->name.nodename,0 ,sizeof (ns->name.nodename)); ret =cr_read_string (ctx, ns->name.nodename, hh->nodename_len);if (ret <0 ) - goto out; + goto out_hbuf;memset (ns->name.domainname,0 ,sizeof (ns->name.domainname)); ret =cr_read_string (ctx, ns->name.domainname, hh->domainname_len); + + out_hbuf: +cr_hbuf_put (ctx,sizeof (*hh)); out: -cr_hbuf_put (ctx,sizeof (*hh)); +up_write (&uts_sem);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
SH> At checkpoint, you only need the uts_sem right around the reading SH> (and length-checking) of uts_ns->name. You're keeping it around SH> writing into the checkpoint image (and cr_hbuf_get).
Right, okay.
SH> At restart, you shouldn't actually need uts_sem - the nsproxy is SH> not yet available to anyone else, so there is no chance of a race, SH> right? Hmm, well maybe I'm being overly optimistic, since restart SH> is coordinated by userspace which could mess up and somehow run SH> sethostname(2) while it's supposed to be running sys_restart(). SH> But then sys_restart() wouldn't yet have hooked up the task to the SH> new nsproxy.
Good, you passed the test. Yeah, that's it... :)
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): 87bpqwa6ku.fsf-FLMGYpZoEPULwtHQx/[email protected]
Quoted 14 lines · Dan Smith
SH> At checkpoint, you only need the
uts_semright around the reading SH> (and length-checking) ofuts_ns->name. You're keeping it around SH> writing into the checkpoint image (andcr_hbuf_get).Right, okay.
SH> At restart, you shouldn't actually need
uts_sem- the nsproxy is SH> not yet available to anyone else, so there is no chance of a race, SH> right? Hmm, well maybe I'm being overly optimistic, since restart SH> is coordinated by userspace which could mess up and somehow run SH> sethostname(2) while it's supposed to be runningsys_restart(). SH> But thensys_restart()wouldn't yet have hooked up the task to the SH> new nsproxy.
Quoted 1 line
Good, you passed the test. Yeah, that's it... :)
woohoo
-serge