Bug#991119: unblock: postsrsd/1.10-2
Control: tags -1 confirmed moreinfo
On 2021-07-14 21:48:50, Oxan van Leeuwen wrote:
> Package: release.debian.org
> Severity: normal
> User: release.debian.org@packages.debian.org
> Usertags: unblock
>
> Please unblock package postsrsd
>
> [ Reason ]
> Security fix for CVE-2021-35525.
>
> [ Impact ]
> Package is vulnerable to a potential DoS attack.
>
> [ Tests ]
> Tests from upstream backported, testsuite from upstream passes, manually tested
> functionality.
>
> [ Risks ]
> Fix is a one-to-one backport from upstream, modulus formatting changes.
>
> [ Checklist ]
> [x] all changes are documented in the d/changelog
> [x] I reviewed all changes and I approve them
> [x] attach debdiff against the package in testing
>
> [ Other info ]
> N/A
>
> unblock postsrsd/1.10-2
If this is a pre-approval request, please go ahead and remove the
moreinfo tag once the new version is available in unstable.
Cheers
> diff -Nru postsrsd-1.10/debian/changelog postsrsd-1.10/debian/changelog
> --- postsrsd-1.10/debian/changelog 2020-12-02 22:36:36.000000000 +0100
> +++ postsrsd-1.10/debian/changelog 2021-07-14 21:21:11.000000000 +0200
> @@ -1,4 +1,12 @@
> -postsrsd (1.10-1) UNRELEASED; urgency=medium
> +postsrsd (1.10-2) UNRELEASED; urgency=medium
> +
> + * Fix CVE-2021-35525: potential DoS when Postfix sends certain long data
> + fields such as multiple concatenated email addresses. Fix backported from
> + upstream commit 077be98d8c8. (Closes: #990439)
> +
> + -- Oxan van Leeuwen <oxan@oxanvanleeuwen.nl> Wed, 14 Jul 2021 21:21:11 +0200
> +
> +postsrsd (1.10-1) unstable; urgency=medium
>
> * New upstream release (Closes: #975633)
> * Drop patches integrated upstream
> diff -Nru postsrsd-1.10/debian/patches/0002-SECURITY-Fix-DoS-on-overly-long-input-from-Postfix.patch postsrsd-1.10/debian/patches/0002-SECURITY-Fix-DoS-on-overly-long-input-from-Postfix.patch
> --- postsrsd-1.10/debian/patches/0002-SECURITY-Fix-DoS-on-overly-long-input-from-Postfix.patch 1970-01-01 01:00:00.000000000 +0100
> +++ postsrsd-1.10/debian/patches/0002-SECURITY-Fix-DoS-on-overly-long-input-from-Postfix.patch 2021-07-14 21:21:11.000000000 +0200
> @@ -0,0 +1,211 @@
> +From: =?utf-8?q?Timo_R=C3=B6hling?= <timo@gaussglocke.de>
> +Date: Sun, 21 Mar 2021 15:27:55 +0100
> +Subject: SECURITY: Fix DoS on overly long input from Postfix
> +MIME-Version: 1.0
> +Content-Type: text/plain; charset="utf-8"
> +Content-Transfer-Encoding: 8bit
> +
> +Thanks to Mateusz Jończyk who reported this issue and gave valuable
> +feedback for its resolution.
> +
> +PostSRSd would hang on an overly long GET request, because the
> +fread()/fwrite() logic in the subprocess would get confused by the
> +remaining input line in its buffer.
> +
> +Theoretically, this error should never occur, as Postfix is supposed to
> +send valid email addresses only, which are shorter than the buffer, even
> +assuming every single character is percent-encoded. However, Postfix
> +sometimes does seem to send malformed request with multiple concatenated
> +email addresses. I'm not sure if there's a reliable way to trigger this
> +condition by an external attacker, but it is a security bug in PostSRSd
> +nevertheless.
> +
> +Fixes CVE-2021-35525.
> +
> +Origin: https://github.com/roehling/postsrsd/commit/077be98d8c8a9847e4ae0c7dc09e7474cbe27db2
> +Forwarded: not-needed
> +Last-Update: 2021-07-14
> +---
> + postsrsd.c | 52 ++++++++++++++++++++++++++++++-------------------
> + run_postsrsd_tests.bats | 40 +++++++++++++++++++++++++++++++++----
> + 2 files changed, 68 insertions(+), 24 deletions(-)
> +
> +diff --git a/postsrsd.c b/postsrsd.c
> +index c009d8f..5ebf7f6 100644
> +--- a/postsrsd.c
> ++++ b/postsrsd.c
> +@@ -518,9 +518,9 @@ int main (int argc, char **argv)
> + fds[sc].events = POLLIN;
> + }
> + while(TRUE) {
> + int conn;
> +- FILE *fp;
> ++ FILE *fp_read, *fp_write;
> + char linebuf[1024], *line;
> + char keybuf[1024], *key;
> +
> + if (poll(fds, socket_count, 1000) < 0) {
> +@@ -540,41 +540,53 @@ int main (int argc, char **argv)
> + int i;
> + // close listen sockets so that we don't stop the main daemon process from restarting
> + for (i = 0; i < socket_count; ++i) close (sockets[i]);
> +
> +- fp = fdopen(conn, "r+");
> +- if (fp == NULL) exit(EXIT_FAILURE);
> +- fds[0].fd = conn;
> +- fds[0].events = POLLIN;
> +- if (poll(fds, 1, timeout * 1000) <= 0) return EXIT_FAILURE;
> +- line = fgets(linebuf, sizeof(linebuf), fp);
> +- while (line) {
> +- fseek (fp, 0, SEEK_CUR); /* Workaround for Solaris */
> ++ /* create separate input/output streams */
> ++ fp_read = fdopen(conn, "r");
> ++ if (fp_read == NULL)
> ++ return EXIT_FAILURE;
> ++ fp_write = fdopen(dup(conn), "w");
> ++ if (fp_write == NULL) return EXIT_FAILURE;
> ++ errno = 0;
> ++ alarm(timeout);
> ++ if (errno != 0)
> ++ return EXIT_FAILURE;
> ++ while ((line = fgets(linebuf, sizeof(linebuf), fp_read))) {
> + char* token;
> ++ alarm(0);
> ++ if (strlen(line) >= sizeof(linebuf) - 1) {
> ++ fprintf(fp_write, "500 Invalid request\n");
> ++ fflush(fp_write);
> ++ return EXIT_FAILURE;
> ++ }
> + token = strtok(line, " \r\n");
> + if (token == NULL || strcmp(token, "get") != 0) {
> +- fprintf (fp, "500 Invalid request\n");
> +- fflush (fp);
> ++ fprintf (fp_write, "500 Invalid request\n");
> ++ fflush (fp_write);
> + return EXIT_FAILURE;
> + }
> + token = strtok(NULL, "\r\n");
> + if (!token) {
> +- fprintf (fp, "500 Invalid request\n");
> +- fflush (fp);
> ++ fprintf (fp_write, "500 Invalid request\n");
> ++ fflush (fp_write);
> + return EXIT_FAILURE;
> + }
> + key = url_decode(keybuf, sizeof(keybuf), token);
> + if (!key) {
> +- fprintf (fp, "500 Invalid request\n");
> +- fflush(fp);
> ++ fprintf (fp_write, "500 Invalid request\n");
> ++ fflush(fp_write);
> + return EXIT_FAILURE;
> + }
> +- handler[sc](srs, fp, key, domain, excludes);
> +- fflush (fp);
> +- if (poll(fds, 1, timeout * 1000) <= 0) break;
> +- line = fgets(linebuf, sizeof(linebuf), fp);
> ++ handler[sc](srs, fp_write, key, domain, excludes);
> ++ fflush (fp_write);
> ++ errno = 0;
> ++ alarm(timeout);
> ++ if (errno != 0)
> ++ return EXIT_FAILURE;
> + }
> +- fclose (fp);
> ++ fclose (fp_write);
> ++ fclose (fp_read);
> + return EXIT_SUCCESS;
> + }
> + close (conn);
> + }
> +diff --git a/run_postsrsd_tests.bats b/run_postsrsd_tests.bats
> +index f4b04bb..3d52a50 100755
> +--- a/run_postsrsd_tests.bats
> ++++ b/run_postsrsd_tests.bats
> +@@ -2,9 +2,9 @@
> + # vim: filetype=bash:
> +
> + if [ ! -x "$POSTSRSD" ]
> + then
> +- for builddir in . build* obj*
> ++ for builddir in . build* obj* _build*
> + do
> + if [ -x "${builddir}/postsrsd" ]
> + then
> + POSTSRSD="${builddir}/postsrsd"
> +@@ -14,9 +14,9 @@ then
> + fi
> + if [ ! -x "$POSTSRSD" ]
> + then
> + cat>&2 <<- EOF
> +- cannot find postsrsd executable (looked in ., build*, obj*)
> ++ cannot find postsrsd executable (looked in ., build*, obj*, _build*)
> + please build the executable first, or set the POSTSRSD
> + environment variable if it is in a different location.
> +
> + EOF
> +@@ -25,14 +25,21 @@ fi
> +
> + LANG=C.UTF-8
> +
> +
> ++fillchar()
> ++{
> ++ local count="$1"
> ++ local char="$2"
> ++ eval 'printf "'"$char"'%.0s" {1..'"$count"'}'
> ++}
> ++
> + start_postsrsd_at()
> + {
> + echo 'tops3cr3t' > "$BATS_TMPDIR/postsrsd.secret"
> + local faketime="$1"
> + shift
> +- faketime "${faketime}" ${POSTSRSD} -D -f 10001 -r 10002 -p "$BATS_TMPDIR/postsrsd.pid" -s "$BATS_TMPDIR/postsrsd.secret" -d example.com "$@"
> ++ faketime "${faketime}" ${POSTSRSD} -D -t1 -f 10001 -r 10002 -p "$BATS_TMPDIR/postsrsd.pid" -s "$BATS_TMPDIR/postsrsd.secret" -d example.com "$@"
> + }
> +
> + stop_postsrsd()
> + {
> +@@ -158,9 +165,9 @@ teardown()
> + read<&9 line
> + [[ "$line" =~ ^"500 Domain excluded" ]]
> + }
> +
> +-@test "SRS invalid requests" {
> ++@test "Malformed or invalid requests" {
> + start_postsrsd_at "2020-01-01 00:01:00 UTC"
> + exec 9<>/dev/tcp/127.0.0.1/10001
> + echo>&9 "get"
> + read<&9 line
> +@@ -172,5 +179,30 @@ teardown()
> + exec 9<>/dev/tcp/127.0.0.1/10001
> + echo>&9 "get encoding%error@otherdomain.com"
> + read<&9 line
> + [[ "$line" =~ ^500 ]]
> ++ exec 9<>/dev/tcp/127.0.0.1/10001
> ++ # Try to overflow the input buffer
> ++ echo>&9 "get too_long@`fillchar 1024 a`.com"
> ++ read<&9 line
> ++ [[ "$line" =~ ^500 ]]
> ++}
> ++
> ++@test "Pipelining multiple requests" {
> ++ start_postsrsd_at "2020-01-01 00:01:00 UTC"
> ++ exec 9<>/dev/tcp/127.0.0.1/10001
> ++ # Send two requests at once and see if PostSRSd answers both
> ++ echo>&9 -e "get test@domain1.com\nget test@domain2.com"
> ++ read<&9 line
> ++ [[ "$line" =~ ^200 ]]
> ++ read<&9 line
> ++ [[ "$line" =~ ^200 ]]
> ++}
> ++
> ++@test "Session timeout" {
> ++ start_postsrsd_at "2020-01-01 00:01:00 UTC"
> ++ exec 9<>/dev/tcp/127.0.0.1/10001
> ++ # Wait until PostSRSd disconnects due to inactivity
> ++ sleep 2
> ++ echo >&9 "get test@example.com"
> ++ ! read <&9 line
> + }
> diff -Nru postsrsd-1.10/debian/patches/series postsrsd-1.10/debian/patches/series
> --- postsrsd-1.10/debian/patches/series 2020-12-02 22:36:36.000000000 +0100
> +++ postsrsd-1.10/debian/patches/series 2021-07-14 21:21:11.000000000 +0200
> @@ -1 +1,2 @@
> 0001-Run-as-postsrsd-user-by-default.patch
> +0002-SECURITY-Fix-DoS-on-overly-long-input-from-Postfix.patch
--
Sebastian Ramacher
Reply to: