new release 3.4.0 - critical security release

rsync.project rsync.project at gmail.com
Wed Jan 15 07:02:16 UTC 2025


I'd also note that the patches for 3.4.0 were made available to 81
different vendors via VINCE since 18th December (under embargo to give time
for vendors to prepare). It is unfortunate that it didn't cover the
platform you are building for. What platform is it btw?


On Wed, 15 Jan 2025 at 17:57, rsync.project <rsync.project at gmail.com> wrote:

> The popt changes came from upstream popt. We have Solaris and FreeBSD CI
> tests, along with linux, but don't have a method for testing other
> platforms. If you submit a PR to fix this, please consider a way we can
> test the fix in CI.
> Cheers, Tridge
>
> On Wed, 15 Jan 2025 at 14:35, <rsbecker at nexbridge.com> wrote:
>
>> Another issue here in findme.c. strlcpy() is a BSD-only method and
>> definitely not portable.
>>
>> Please consider other platforms when creating patches. I can provide a
>> patch to this
>>
>> patch also.
>>
>>
>>
>> Thanks,
>>
>> Randall
>>
>>
>>
>> *From:* rsync <rsync-bounces at lists.samba.org> *On Behalf Of *Randall S.
>> Becker via rsync
>> *Sent:* January 14, 2025 6:46 PM
>> *To:* 'rsync.project' <rsync.project at gmail.com>
>> *Cc:* rsync at lists.samba.org
>> *Subject:* RE: new release 3.4.0 - critical security release
>>
>>
>>
>> Here is my fix for the situation:
>>
>>
>>
>> *diff --git a/popt/findme.c b/popt/findme.c*
>>
>> *index ac4cbae..4fe8a18 100644*
>>
>> *--- a/popt/findme.c*
>>
>> *+++ b/popt/findme.c*
>>
>> @@ -25,12 +25,23 @@ const char * findProgramPath(const char * argv0)
>>
>>      if (path == NULL) return NULL;
>>
>>
>>
>>      bufsize = strlen(path) + 1;
>>
>> +#if defined __TANDEM
>>
>> +    start = pathbuf = malloc(bufsize);
>>
>> +#else
>>
>>      start = pathbuf = alloca(bufsize);
>>
>> +#endif
>>
>>      if (pathbuf == NULL) return NULL;  /* XXX can't happen */
>>
>>      strlcpy(pathbuf, path, bufsize);
>>
>>      bufsize += sizeof "/" - 1 + strlen(argv0);
>>
>>      buf = malloc(bufsize);
>>
>> +#if defined __TANDEM
>>
>> +    if (buf == NULL) {
>>
>> +       free(start);
>>
>> +       return NULL;    /* XXX can't happen */
>>
>> +    }
>>
>> +#else
>>
>>      if (buf == NULL) return NULL;      /* XXX can't happen */
>>
>> +#endif
>>
>>
>>
>>      chptr = NULL;
>>
>>      /*@-branchstate@*/
>>
>> @@ -39,8 +50,15 @@ const char * findProgramPath(const char * argv0)
>>
>>             *chptr = '\0';
>>
>>         snprintf(buf, bufsize, "%s/%s", start, argv0);
>>
>>
>>
>> +#if defined __TANDEM
>>
>> +       if (!access(buf, X_OK)) {
>>
>> +           free(start);
>>
>> +           return buf;
>>
>> +       }
>>
>> +#else
>>
>>         if (!access(buf, X_OK))
>>
>>             return buf;
>>
>> +#endif
>>
>>
>>
>>         if (chptr)
>>
>>             start = chptr + 1;
>>
>> @@ -51,5 +69,8 @@ const char * findProgramPath(const char * argv0)
>>
>>
>>
>>      free(buf);
>>
>>
>>
>> +#if defined __TANDEM
>>
>> +    free(start);
>>
>> +#endif
>>
>>      return NULL;
>>
>> }
>>
>>
>>
>> I would respectfully ask that it be included ASAP.
>>
>>
>>
>> Thanks,
>>
>> Randall
>>
>>
>>
>> *From:* rsync <rsync-bounces at lists.samba.org> *On Behalf Of *Randall S.
>> Becker via rsync
>> *Sent:* January 14, 2025 6:09 PM
>> *To:* 'rsync.project' <rsync.project at gmail.com>
>> *Cc:* rsync at lists.samba.org
>> *Subject:* RE: new release 3.4.0 - critical security release
>>
>>
>>
>> This happens on NonStop x86 and ia64. I have been building/packaging
>> Rsync for years – almost a decade in fact. I think this happened once
>> before this year, in fact.
>>
>>
>>
>> It is equivalent to the more portable malloc/free, which I would prefer
>> to have in this series even if it has to be wrapped in a #if defined
>> (__TANDEM) block.
>>
>>
>>
>> This call is considered not portable and allocates on the stack instead
>> of the heap. This can cause performance issues as memory management on the
>> heap is generally given more attention by runtimes. The reason it is not
>> supported on NonStop is that the c99 compiler does not generate code for
>> allocating on the stack on this machine.
>>
>>
>>
>> Please forgive me here, but adding a new dependency for a critical
>> security fix is rather painful.
>>
>>
>>
>> --Randall
>>
>>
>>
>>
>>
>> *From:* rsync.project <rsync.project at gmail.com>
>> *Sent:* January 14, 2025 4:31 PM
>> *To:* rsbecker at nexbridge.com
>> *Cc:* rsync at lists.samba.org
>> *Subject:* Re: new release 3.4.0 - critical security release
>>
>>
>>
>> the alloca comes from the new popt release. What system are you having an
>> issue with?
>>
>>
>>
>>
>>
>> On Wed, 15 Jan 2025 at 07:16, <rsbecker at nexbridge.com> wrote:
>>
>> A new dependency was added since 3.3, alloca(), which is not portable. Is
>> there a way around this?
>>
>> Thanks,
>>
>> Randall
>>
>>
>>
>> *From:* rsync <rsync-bounces at lists.samba.org> *On Behalf Of *rsync.project
>> via rsync
>> *Sent:* January 14, 2025 2:49 PM
>> *To:* rsync-announce at lists.samba.org
>> *Cc:* rsync at lists.samba.org
>> *Subject:* new release 3.4.0 - critical security release
>>
>>
>>
>> We have just released version 3.4.0 of rsync. This release fixes 6
>> security vulnerabilities found by two groups of security researchers.
>>
>>
>>
>> You can find the new release links here:
>>
>>
>>
>>  - https://rsync.samba.org/
>>
>>  - https://download.samba.org/pub/rsync/src/
>>
>>
>>
>> For details on the vulnerabilities please see this CERT advisory:
>>
>>
>>
>> https://kb.cert.org/vuls/id/952657
>>
>>
>>
>> The various distros should be doing security releases today
>>
>> Many thanks to Simon Scannell, Pedro Gallegos, and Jasiel Spelman at
>> Google Cloud Vulnerability Research and Aleksei Gorban (Loqpa) for
>> discovering these vulnerabilities and working with the rsync project to
>> develop and test fixes.
>>
>>
>>
>> Also many thanks to Wayne Davison for assisting with the release process
>> as this is the first release I've done since 2002 when Wayne took over as
>> the rsync maintainer.
>>
>>
>>
>> Andrew Tridgell
>>
>> rsync maintainer (again!)
>>
>>
>>
>>
>>
>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.samba.org/pipermail/rsync/attachments/20250115/1adf349d/attachment-0001.htm>


More information about the rsync mailing list