On Thu, 2017-06-08 at 15:14 +0200, Paul Menzel wrote:
> Dear autofs folks,
>
>
> Valgrind 3.12.0 shows the following when running `automount`.
>
> ```
> ==63563== Thread 12:
> ==63563== Conditional jump or move depends on uninitialised value(s)
> ==63563== at 0x4C2E826: rawmemchr (vg_replace_strmem.c:1402)
> ==63563== by 0x61D3FB6: _nss_files_parse_grent (fgetgrent_r.c:37)
> ==63563== by 0xB3CA5FA: internal_getent (files-XXX.c:272)
> ==63563== by 0xB3CA9E0: _nss_files_getgrgid_r (files-grp.c:39)
> ==63563== by 0x61D393B: getgrgid_r@@GLIBC_2.2.5 (getXXbyYY_r.c:266)
> ==63563== by 0x131AC3: set_tsd_user_vars (mounts.c:1538)
> ==63563== by 0x11B652: do_mount_indirect (indirect.c:776)
> ==63563== by 0x4E3B190: start_thread (pthread_create.c:309)
> ==63563== Uninitialised value was created by a heap allocation
> ==63563== at 0x4C27A10: malloc (vg_replace_malloc.c:298)
> ==63563== by 0x4C299DC: realloc (vg_replace_malloc.c:785)
> ==63563== by 0x131A4D: set_tsd_user_vars (mounts.c:1530)
> ==63563== by 0x11B652: do_mount_indirect (indirect.c:776)
> ==63563== by 0x4E3B190: start_thread (pthread_create.c:309)
> ==63563==
> ==63563== Conditional jump or move depends on uninitialised value(s)
> ==63563== at 0x4C2A4A5: __GI_strchr (vg_replace_strmem.c:246)
> ==63563== by 0x61D3DBB: _nss_files_parse_grent (fgetgrent_r.c:37)
> ==63563== by 0xB3CA5FA: internal_getent (files-XXX.c:272)
> ==63563== by 0xB3CA9E0: _nss_files_getgrgid_r (files-grp.c:39)
> ==63563== by 0x61D393B: getgrgid_r@@GLIBC_2.2.5 (getXXbyYY_r.c:266)
> ==63563== by 0x131AC3: set_tsd_user_vars (mounts.c:1538)
> ==63563== by 0x11B652: do_mount_indirect (indirect.c:776)
> ==63563== by 0x4E3B190: start_thread (pthread_create.c:309)
> ==63563== Uninitialised value was created by a heap allocation
> ==63563== at 0x4C27A10: malloc (vg_replace_malloc.c:298)
> ==63563== by 0x4C299DC: realloc (vg_replace_malloc.c:785)
> ==63563== by 0x131A4D: set_tsd_user_vars (mounts.c:1530)
> ==63563== by 0x11B652: do_mount_indirect (indirect.c:776)
> ==63563== by 0x4E3B190: start_thread (pthread_create.c:309)
> ==63563==
> ==63563== Conditional jump or move depends on uninitialised value(s)
> ==63563== at 0x4C2A49A: __GI_strchr (vg_replace_strmem.c:246)
> ==63563== by 0x61D3DBB: _nss_files_parse_grent (fgetgrent_r.c:37)
> ==63563== by 0xB3CA5FA: internal_getent (files-XXX.c:272)
> ==63563== by 0xB3CA9E0: _nss_files_getgrgid_r (files-grp.c:39)
> ==63563== by 0x61D393B: getgrgid_r@@GLIBC_2.2.5 (getXXbyYY_r.c:266)
> ==63563== by 0x131AC3: set_tsd_user_vars (mounts.c:1538)
> ==63563== by 0x11B652: do_mount_indirect (indirect.c:776)
> ==63563== by 0x4E3B190: start_thread (pthread_create.c:309)
> ==63563== Uninitialised value was created by a heap allocation
> ==63563== at 0x4C27A10: malloc (vg_replace_malloc.c:298)
> ==63563== by 0x4C299DC: realloc (vg_replace_malloc.c:785)
> ==63563== by 0x131A4D: set_tsd_user_vars (mounts.c:1530)
> ==63563== by 0x11B652: do_mount_indirect (indirect.c:776)
> ==63563== by 0x4E3B190: start_thread (pthread_create.c:309)
> ==63563==
> ```
>
> The corresponding code looks like below.
>
> ```
> gr_tmp = NULL;
> status = ERANGE;
> #ifdef ENABLE_LIMIT_GETGRGID_SIZE
> if (!maxgrpbuf)
> maxgrpbuf = detached_thread_stack_size * 0.9;
> #endif
>
> /* If getting the group name fails go on without it. It's
> * used to set an environment variable for program maps
> * which may or may not use it so it isn't critical to
> * operation.
> */
>
> tmplen = grplen;
> while (1) {
> char *tmp = realloc(gr_tmp, tmplen + 1);
> if (!tmp) {
> error(logopt, "failed to malloc buffer for
> getgrgid_r");
> goto no_group;
> }
> gr_tmp = tmp;
> pgr = &gr;
> ppgr = &pgr;
> status = getgrgid_r(gid, pgr, gr_tmp, tmplen, ppgr);
> if (status != ERANGE)
> break;
> tmplen += grplen;
>
> /* Don't tempt glibc to alloca() larger than is (likely)
> * available on the stack if limit-getgrgid-size is
> enabled.
> */
> if (!maxgrpbuf || (tmplen < maxgrpbuf))
> continue;
>
> /* Add a message so we know this happened */
> debug(logopt, "group buffer allocation would be too
> large");
> break;
> }
> ```
>
> Unfortunately, I don’t spot the problem. Do you see it?
I've looked at that area quite a bit recently and looking again I don't see what
it's talking about either.
Puzzling.
Ian
--
To unsubscribe from this list: send the line "unsubscribe autofs" in
