Compare commits

...

No commits in common. "c8" and "c8-beta" have entirely different histories.
c8 ... c8-beta

7 changed files with 1 additions and 853 deletions

View File

@ -1,46 +0,0 @@
From d2f39a20d68aa55023b63e575f06941721e644c6 Mon Sep 17 00:00:00 2001
From: Paulo Alcantara <pc@manguebit.org>
Date: Tue, 23 Jun 2026 15:06:20 -0300
Subject: [PATCH] cifs.upcall: fix compiler warning with -Wvla
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
The length value for @path array in get_uidgid() needs to be evaluated
at compile time, so replace strlen() with sizeof() when defining
PROC_PID_PATH_MAXLEN and then fix the following warning:
cifs.upcall.c: In function get_uidgid:
cifs.upcall.c:1400:9: warning: ISO C90 forbids array path whose size
cannot be evaluated [-Wvla]
1400 | char path[PROC_PID_PATH_MAXLEN] = {}, buf[256];
| ^~~~
Fixes: 972c5b5ff95e ("cifs.upcall: remove getpwuid() dependency")
Signed-off-by: Paulo Alcantara (Red Hat) <pc@manguebit.org>
Reviewed-by: David Howells <dhowells@redhat.com>
Cc: Enzo Matsumiya <ematsumiya@suse.de>
Cc: linux-cifs@vger.kernel.org
Signed-off-by: Steve French <stfrench@microsoft.com>
---
cifs.upcall.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/cifs.upcall.c b/cifs.upcall.c
index 01690dfcade1..11dbc6186a74 100644
--- a/cifs.upcall.c
+++ b/cifs.upcall.c
@@ -1375,8 +1375,8 @@ static int ip_to_fqdn(const char *addrstr, char *host, size_t hostlen)
return 0;
}
-/* cover worst case/impossible scenarios, + 1 for NUL */
-#define PROC_PID_PATH_MAXLEN ((int)strlen("/proc/2147483647/status") + 1)
+/* cover worst case/impossible scenarios */
+#define PROC_PID_PATH_MAXLEN ((int)sizeof("/proc/2147483647/status"))
/* max valid UID/GID is (UINT_MAX - 1) */
#define INVALID_UIDGID UINT_MAX
--
2.54.0

View File

@ -1,383 +0,0 @@
From e9495963e0d5c26f7d0137829d8dc625130b53cc Mon Sep 17 00:00:00 2001
From: Enzo Matsumiya <ematsumiya@suse.de>
Date: Mon, 6 Jul 2026 10:55:55 -0300
Subject: [PATCH] cifs.upcall: fix regression with krb5 + creduid
Commit 972c5b5ff95e ("cifs.upcall: remove getpwuid() dependency")
introduced a regression when using creduid != uid (e.g.
"mount.cifs -o sec=krb5,cruid=X"), so 'uid' local var is replaced
with procfs "Uid" value (in the example, the one from mount.cifs).
That commit ignored the fact that:
mount UID can be different from creds UID, and that calling-app
process (post-mount) can be different from both
This patch "reverts" 972c5b5ff95e ("cifs.upcall: remove getpwuid()
dependency"); transform the "emergency"-added function get_uidgid()
into map_uidgid(), now only used to do NS UID/GID mapping.
Also add getpwuid() back, but this time called while still on host
namespace, so any possible custom NSS module is ran as allowed by
sysadmin.
Any scenario involving unmapped UIDs or GIDs is unsupported; this
means that any UID:GID in a child user namespace _must_ map back to
a valid and existing host UID:GID.
Fixes: 972c5b5ff95e ("cifs.upcall: remove getpwuid() dependency")
Reported-by: Paulo Alcantara (Red Hat) <pc@manguebit.org>
Signed-off-by: Enzo Matsumiya <ematsumiya@suse.de>
Signed-off-by: Paulo Alcantara (Red Hat) <pc@manguebit.org>
Signed-off-by: Steve French <stfrench@microsoft.com>
---
cifs.upcall.c | 279 ++++++++++++++++++++++++++++++++------------------
1 file changed, 177 insertions(+), 102 deletions(-)
Index: cifs-utils-7.0/cifs.upcall.c
===================================================================
--- cifs-utils-7.0.orig/cifs.upcall.c
+++ cifs-utils-7.0/cifs.upcall.c
@@ -52,6 +52,7 @@
#include <arpa/inet.h>
#include <ctype.h>
#include <grp.h>
+#include <pwd.h>
#include <stdbool.h>
#include <errno.h>
#include <sched.h>
@@ -1264,118 +1265,170 @@ static int ip_to_fqdn(const char *addrst
return 0;
}
-/* cover worst case/impossible scenarios */
-#define PROC_PID_PATH_MAXLEN ((int)sizeof("/proc/2147483647/status"))
-/* max valid UID/GID is (UINT_MAX - 1) */
-#define INVALID_UIDGID UINT_MAX
+#define ID_MAP_PATH_MAX ((int)sizeof("/proc/2147483647/uid_map"))
-/*
- * get_uidgid - Get @pid's (real) UID and/or GID.
- * @pid: process to get UID/GID from
- * @uidp: pointer to store @pid's UID (can be NULL)
- * @gidp: pointer to store @pid's GID (can be NULL)
- *
- * Extract "Uid:" and "Gid:" fields from /proc/@pid/status.
- * Do so based on whether @uidp or @gidp are NULL.
- *
- * This function assumes we're on the same namespace as @pid.
- *
- * Return: 0 on success, -1 otherwise (errno set).
- *
- * On errors, *@uidp and *@gidp are set to INVALID_UIDGID.
- */
-static int get_uidgid(pid_t pid, uid_t *uidp, gid_t *gidp)
+static int map_id(pid_t pid, const char *map, unsigned int *idp)
{
- char path[PROC_PID_PATH_MAXLEN] = {}, buf[256];
+ unsigned long long ns_start, host_start, range;
+ char map_path[ID_MAP_PATH_MAX];
+ int map_path_size = sizeof(map_path);
+ unsigned int id;
FILE *fp = NULL;
- int ret;
+ int ret = 1;
- errno = 0;
- if (pid < 0 || (!uidp && !gidp)) {
- errno = EINVAL;
- return -1;
- }
-
- if (uidp)
- *uidp = INVALID_UIDGID;
-
- if (gidp)
- *gidp = INVALID_UIDGID;
+ errno = EINVAL;
+ if (pid < 0 || !map || !idp || *idp == UINT_MAX)
+ goto out;
- ret = snprintf(path, PROC_PID_PATH_MAXLEN, "/proc/%d/status", pid);
- if (ret < 0 || ret >= PROC_PID_PATH_MAXLEN) {
- if (!errno)
+ ret = snprintf(map_path, map_path_size, "/proc/%d/%s", pid, map);
+ if (ret < 0 || ret >= map_path_size) {
+ if (ret >= map_path_size)
errno = ENAMETOOLONG;
- return -1;
+ ret = 1;
+ goto out;
}
- fp = fopen(path, "r");
- if (!fp) {
- ret = -1;
+ ret = 1;
+ fp = fopen(map_path, "r");
+ if (!fp)
goto out;
- }
- /* Parse /proc/pid/status fields */
- errno = 0;
- ret = -1;
- while (fgets(buf, 256, fp)) {
- unsigned long long val;
+ /*
+ * The map files have the same format:
+ * <NS ID start> <host ID start> <range>
+ * ... (<multiple entries are supported) ...
+ *
+ * Formula and validation:
+ * <final NS ID> = (*@idp - <host ID start>) + <NS ID start>
+ *
+ * - IDs: [0, UINT_MAX - 1]
+ * - range: [1, UINT_MAX], where range == UINT_MAX requires both ID ranges to start at 0,
+ * which then means this is an init host NS mapping (and that's ok)
+ *
+ * The formula itself would be enough to "validate" a matching NS ID, but we don't want to
+ * keep parsing a malformed map file, no matter how unlikely/impossible it is to happen.
+ * Same reason values are parsed as 'unsigned long long', so we can check for bogus data.
+ */
+ id = UINT_MAX;
+ errno = ENODATA;
+ while (fscanf(fp, "%llu %llu %llu", &ns_start, &host_start, &range) == 3) {
+ if (ns_start >= UINT_MAX || host_start >= UINT_MAX ||
+ range > UINT_MAX || range == 0) {
+ errno = EINVAL;
+ ret = 1;
+ break;
+ }
- errno = ENODATA;
- if ((!uidp || strncmp(buf, "Uid:", 4)) && (!gidp || strncmp(buf, "Gid:", 4)))
- continue;
+ if (range == UINT_MAX && (ns_start != 0 || host_start != 0)) {
+ errno = EINVAL;
+ ret = 1;
+ break;
+ }
- errno = 0;
+ if (host_start + range > UINT_MAX || ns_start + range > UINT_MAX) {
+ errno = EINVAL;
+ ret = 1;
+ break;
+ }
/*
- * Example line format (same for both Uid/Gid):
- * "Uid:\t%u\t%u\%u\%u"
+ * Check if host ID fits this line.
+ * Our desired NS ID may be in any line of the file.
*
- * Where the numbers represents:
- * <real> <effective> <saved> <fsuid>
+ * Note: new{uid,gid}map tools (that creates the map files) don't allow multiple
+ * maps (NS IDs) to the same host ID.
*
- * We're only interested in the <real> value.
+ * If we get a match here, we'll save it, but we continue parsing the file.
+ * If we happen to find a duplicate, it's possible this is a rogue file trying to
+ * bypass these checks.
*
- * (field names "Uid:"/"Gid:" parsed above, skip it)
+ * In such cases, discard the match and return EOPNOTSUPP, as returning a
+ * successful match could lead to disastrous results.
*/
- ret = sscanf(&buf[0] + 4, "%llu", &val);
- if (ret != 1) {
- ret = -1;
- if (errno)
+ if (*idp >= host_start && *idp < host_start + range) {
+ /* This means we found a duplicate */
+ if (!ret) {
+ ret = 1;
+ errno = EOPNOTSUPP;
+ syslog(LOG_ERR, "%s has multiple mapped IDs for %u (unsupported)",
+ map, *idp);
break;
- continue;
- }
-
- ret = -1;
- if (val >= UINT_MAX) {
- errno = EINVAL;
- break;
- }
-
- if (uidp && !strncmp(buf, "Uid:", 4))
- *uidp = (uid_t)val;
- else
- *gidp = (gid_t)val;
+ }
- if ((!uidp || *uidp != INVALID_UIDGID) && (!gidp || *gidp != INVALID_UIDGID)) {
- errno = 0;
+ id = (*idp - host_start) + ns_start;
ret = 0;
- break;
+ errno = 0;
}
}
+
+ /* This means errno was reset by fscanf() without finding anything */
+ if (ret && errno == 0)
+ errno = ENODATA;
out:
- if (fp)
+ if (fp) {
+ int err = errno;
+
fclose(fp);
+ /* Ignore fclose() errors */
+ errno = err;
+ }
- if (ret) {
- syslog(LOG_DEBUG, "%s(pid=%d): %s", __func__, pid, strerror(errno));
- if (uidp)
- *uidp = INVALID_UIDGID;
+ if (!ret) {
+ *idp = id;
+ errno = 0;
+ } else {
+ syslog(LOG_DEBUG, "%s(pid=%d, map=%s, id=%u): %s", __func__, pid, map, *idp,
+ strerror(errno));
+ }
+
+ return ret;
+}
+
+/*
+ * map_uidgid() - Map (real) UID/GID from init host NS to user NS.
+ * @pid: host NS PID
+ * @uidp: (in) host UID, (out) NS UID
+ * @gidp: (in) host GID, (out) NS GID
+ *
+ * Parse /proc/@pid/{uid,gid}_map files to get NS UID/GID values.
+ * Since @pid is expected to be a host NS PID, this must be called before switching namespaces.
+ *
+ * Note: we can't use /proc/self here because we haven't switched NS yet, so {uid,gid}_map files
+ * would contain host NS values.
+ *
+ * Return: 0 on success, 1 otherwise (errno set).
+ */
+static int map_uidgid(pid_t pid, uid_t *uidp, gid_t *gidp)
+{
+ uid_t orig_uid;
+ gid_t orig_gid;
+ int ret = 1;
+
+ errno = EINVAL;
+ if (!uidp || !gidp)
+ goto out;
+
+ orig_uid = *uidp;
+ orig_gid = *gidp;
- if (gidp)
- *gidp = INVALID_UIDGID;
+ ret = map_id(pid, "uid_map", uidp);
+ if (ret) {
+ if (errno == ENODATA)
+ syslog(LOG_ERR, "UID %u not mapped in this namespace (unsupported)",
+ orig_uid);
+ goto out;
}
+ ret = map_id(pid, "gid_map", gidp);
+ if (ret && errno == ENODATA)
+ syslog(LOG_ERR, "GID %u not mapped in this namespace (unsupported)",
+ orig_gid);
+out:
+ if (ret && errno != ENODATA)
+ syslog(LOG_ERR, "%s: %s", __func__, strerror(errno));
+ else if (!ret)
+ syslog(LOG_DEBUG, "host %u:%u -> NS %u:%u", orig_uid, orig_gid, *uidp, *gidp);
+
return ret;
}
@@ -1421,6 +1474,7 @@ int main(const int argc, char *const arg
const char *oid;
uid_t uid;
gid_t gid;
+ struct passwd *pw;
char *keytab_name = NULL;
char *env_cachename = NULL;
krb5_ccache ccache = NULL;
@@ -1561,39 +1615,60 @@ int main(const int argc, char *const arg
}
/*
- * Change to the process's namespace. This means that things will work
- * acceptably in containers, because we'll be looking at the correct
- * filesystem and have the correct network configuration.
+ * 'uid' always points to a host UID, so we must get the corresponding host GID.
+ * It's safe to call getpwuid() here because we're still on host NS, i.e. caller
+ * application has no control over custom NSS modules.
+ *
+ * FIXME: if UID is from another NS, or a subuid, this will fail on mount.
*/
- rc = switch_to_process_ns(arg->pid);
- if (rc == -1) {
- syslog(LOG_ERR, "unable to switch to process namespace: %s", strerror(errno));
+ errno = 0;
+ pw = getpwuid(uid);
+ if (!pw) {
+ syslog(LOG_ERR, "failed to retrieve GID from UID %u: %s", uid,
+ strerror(errno ? errno : ENOENT));
rc = 1;
goto out;
}
- if (trim_capabilities(env_probe))
- goto out;
+ gid = pw->pw_gid;
/*
* We can't reasonably do this for root. When mounting a DFS share,
* for instance we can end up with creds being overridden, but the env
* variable left intact.
+ *
+ * Always check this before NS UID mapping.
*/
if (uid == 0)
env_probe = false;
/*
- * FIXME: this only works if we haven't switched PID namespaces.
- * If we did, /proc/arg->pid/ might not exist, or worse, point to something else.
+ * Map host 'uid' and 'gid' to the target user NS.
+ *
+ * Any scenario that involves unmapped UIDs or primary GIDs is not supported -- we
+ * don't have, and can't find, all the info that would be necessary to find a
+ * UID/GID within all the possible NS combinations.
*/
- rc = get_uidgid(arg->pid, &uid, &gid);
- if (rc) {
- syslog(LOG_ERR, "get_uidgid (NS): %s", strerror(errno));
+ if (!in_same_user_ns(arg->pid, getpid())) {
+ rc = map_uidgid(arg->pid, &uid, &gid);
+ if (rc)
+ goto out;
+ }
+ /*
+ * Change to the process's namespace. This means that things will work
+ * acceptably in containers, because we'll be looking at the correct
+ * filesystem and have the correct network configuration.
+ */
+ rc = switch_to_process_ns(arg->pid);
+ if (rc == -1) {
+ syslog(LOG_ERR, "unable to switch to process namespace: %s", strerror(errno));
rc = 1;
goto out;
}
+ if (trim_capabilities(env_probe))
+ goto out;
+
rc = setgid(gid);
if (rc) {
syslog(LOG_ERR, "setgid: %s", strerror(errno));

View File

@ -1,251 +0,0 @@
From 972c5b5ff95e3e812bc8daa72d0383654ab0dba7 Mon Sep 17 00:00:00 2001
From: Enzo Matsumiya <ematsumiya@suse.de>
Date: Mon, 15 Jun 2026 14:25:22 -0300
Subject: [PATCH] cifs.upcall: remove getpwuid() dependency
This patch removes getpwuid() and resort to getting UID/GID from
/proc/pid/status
Changes:
- add get_uidgid() helper
- drop supplementary groups (i.e. move setgroups(0, NULL)) before even
decoding spnego key -- those are never needed in any case
Signed-off-by: Enzo Matsumiya <ematsumiya@suse.de>
Reported-by: Shaomin Chen <eeesssooo020@gmail.com>
Acked-by: Paulo Alcantara (Red Hat) <pc@manguebit.org>
Acked-by: Pavel Shilovskiy <pshilovskiy@microsoft.com>
Reviewed-by: Shyam Prasad N <sprasad@microsoft.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
---
cifs.upcall.c | 180 ++++++++++++++++++++++++++++++++++++++++----------
1 file changed, 145 insertions(+), 35 deletions(-)
diff --git a/cifs.upcall.c b/cifs.upcall.c
index 52c03280dbe0..ae3294dd61f0 100644
--- a/cifs.upcall.c
+++ b/cifs.upcall.c
@@ -51,7 +51,6 @@
#include <netdb.h>
#include <arpa/inet.h>
#include <ctype.h>
-#include <pwd.h>
#include <grp.h>
#include <stdbool.h>
#include <errno.h>
@@ -1265,6 +1264,121 @@ static int ip_to_fqdn(const char *addrstr, char *host, size_t hostlen)
return 0;
}
+/* cover worst case/impossible scenarios, + 1 for NUL */
+#define PROC_PID_PATH_MAXLEN ((int)strlen("/proc/2147483647/status") + 1)
+/* max valid UID/GID is (UINT_MAX - 1) */
+#define INVALID_UIDGID UINT_MAX
+
+/*
+ * get_uidgid - Get @pid's (real) UID and/or GID.
+ * @pid: process to get UID/GID from
+ * @uidp: pointer to store @pid's UID (can be NULL)
+ * @gidp: pointer to store @pid's GID (can be NULL)
+ *
+ * Extract "Uid:" and "Gid:" fields from /proc/@pid/status.
+ * Do so based on whether @uidp or @gidp are NULL.
+ *
+ * This function assumes we're on the same namespace as @pid.
+ *
+ * Return: 0 on success, -1 otherwise (errno set).
+ *
+ * On errors, *@uidp and *@gidp are set to INVALID_UIDGID.
+ */
+static int get_uidgid(pid_t pid, uid_t *uidp, gid_t *gidp)
+{
+ char path[PROC_PID_PATH_MAXLEN] = {}, buf[256];
+ FILE *fp = NULL;
+ int ret;
+
+ errno = 0;
+ if (pid < 0 || (!uidp && !gidp)) {
+ errno = EINVAL;
+ return -1;
+ }
+
+ if (uidp)
+ *uidp = INVALID_UIDGID;
+
+ if (gidp)
+ *gidp = INVALID_UIDGID;
+
+ ret = snprintf(path, PROC_PID_PATH_MAXLEN, "/proc/%d/status", pid);
+ if (ret < 0 || ret >= PROC_PID_PATH_MAXLEN) {
+ if (!errno)
+ errno = ENAMETOOLONG;
+ return -1;
+ }
+
+ fp = fopen(path, "r");
+ if (!fp) {
+ ret = -1;
+ goto out;
+ }
+
+ /* Parse /proc/pid/status fields */
+ errno = 0;
+ ret = -1;
+ while (fgets(buf, 256, fp)) {
+ unsigned long long val;
+
+ errno = ENODATA;
+ if ((!uidp || strncmp(buf, "Uid:", 4)) && (!gidp || strncmp(buf, "Gid:", 4)))
+ continue;
+
+ errno = 0;
+
+ /*
+ * Example line format (same for both Uid/Gid):
+ * "Uid:\t%u\t%u\%u\%u"
+ *
+ * Where the numbers represents:
+ * <real> <effective> <saved> <fsuid>
+ *
+ * We're only interested in the <real> value.
+ *
+ * (field names "Uid:"/"Gid:" parsed above, skip it)
+ */
+ ret = sscanf(&buf[0] + 4, "%llu", &val);
+ if (ret != 1) {
+ ret = -1;
+ if (errno)
+ break;
+ continue;
+ }
+
+ ret = -1;
+ if (val >= UINT_MAX) {
+ errno = EINVAL;
+ break;
+ }
+
+ if (uidp && !strncmp(buf, "Uid:", 4))
+ *uidp = (uid_t)val;
+ else
+ *gidp = (gid_t)val;
+
+ if ((!uidp || *uidp != INVALID_UIDGID) && (!gidp || *gidp != INVALID_UIDGID)) {
+ errno = 0;
+ ret = 0;
+ break;
+ }
+ }
+out:
+ if (fp)
+ fclose(fp);
+
+ if (ret) {
+ syslog(LOG_DEBUG, "%s(pid=%d): %s", __func__, pid, strerror(errno));
+ if (uidp)
+ *uidp = INVALID_UIDGID;
+
+ if (gidp)
+ *gidp = INVALID_UIDGID;
+ }
+
+ return ret;
+}
+
/* walk a string and lowercase it in-place */
static void
lowercase_string(char *c)
@@ -1306,10 +1420,10 @@ int main(const int argc, char *const argv[])
struct decoded_args *arg = NULL;
const char *oid;
uid_t uid;
+ gid_t gid;
char *keytab_name = NULL;
char *env_cachename = NULL;
krb5_ccache ccache = NULL;
- struct passwd *pw;
unsigned expire_time = DNS_RESOLVER_DEFAULT_TIMEOUT;
const char *key_descr = NULL;
@@ -1396,6 +1510,17 @@ int main(const int argc, char *const argv[])
* Otherwise, it's a spnego key request
*/
+ /*
+ * We don't need supplementary groups, ever.
+ * Drop them ASAP.
+ */
+ rc = setgroups(0, NULL);
+ if (rc == -1) {
+ syslog(LOG_ERR, "setgroups: %s", strerror(errno));
+ rc = 1;
+ goto out;
+ }
+
rc = decode_key_description(buf, &arg);
free(buf);
if (rc) {
@@ -1450,38 +1575,6 @@ int main(const int argc, char *const argv[])
if (trim_capabilities(env_probe))
goto out;
- /*
- * The kernel doesn't pass down the gid, so we resort here to scraping
- * one out of the passwd nss db. Note that this might not reflect the
- * actual gid of the process that initiated the upcall. While we could
- * scrape that out of /proc, relying on that is a bit more risky.
- */
- pw = getpwuid(uid);
- if (!pw) {
- syslog(LOG_ERR, "Unable to find pw entry for uid %d: %s\n",
- uid, strerror(errno));
- rc = 1;
- goto out;
- }
-
- /*
- * The kernel should send down a zero-length grouplist already, but
- * just to be on the safe side...
- */
- rc = setgroups(0, NULL);
- if (rc == -1) {
- syslog(LOG_ERR, "setgroups: %s", strerror(errno));
- rc = 1;
- goto out;
- }
-
- rc = setgid(pw->pw_gid);
- if (rc == -1) {
- syslog(LOG_ERR, "setgid: %s", strerror(errno));
- rc = 1;
- goto out;
- }
-
/*
* We can't reasonably do this for root. When mounting a DFS share,
* for instance we can end up with creds being overridden, but the env
@@ -1490,6 +1583,24 @@ int main(const int argc, char *const argv[])
if (uid == 0)
env_probe = false;
+ /*
+ * FIXME: this only works if we haven't switched PID namespaces.
+ * If we did, /proc/arg->pid/ might not exist, or worse, point to something else.
+ */
+ rc = get_uidgid(arg->pid, &uid, &gid);
+ if (rc) {
+ syslog(LOG_ERR, "get_uidgid (NS): %s", strerror(errno));
+ rc = 1;
+ goto out;
+ }
+
+ rc = setgid(gid);
+ if (rc) {
+ syslog(LOG_ERR, "setgid: %s", strerror(errno));
+ rc = 1;
+ goto out;
+ }
+
/*
* Must do this before setuid, as we need elevated capabilities to
* look at the environ file.

View File

@ -1,35 +0,0 @@
From 4d87803f386764ed0b5949ccf70fc684b744b83a Mon Sep 17 00:00:00 2001
From: Paulo Alcantara <pc@manguebit.org>
Date: Tue, 25 Nov 2025 17:58:29 -0300
Subject: [PATCH] docs: update echo_interval description
It is '3 * echo_interval' since upstream commit f2caf901c1b7 ("cifs:
Fix a race condition with cifs_echo_request").
Reported-by: Alexandros Panagiotou <apanagio@redhat.com>
Signed-off-by: Paulo Alcantara (Red Hat) <pc@manguebit.org>
Cc: Steve French <smfrench@gmail.com>
Cc: linux-cifs@vger.kernel.org
Signed-off-by: Steve French <stfrench@microsoft.com>
---
mount.cifs.rst | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/mount.cifs.rst b/mount.cifs.rst
index 9eee7d5afc06..52fe492a5d37 100644
--- a/mount.cifs.rst
+++ b/mount.cifs.rst
@@ -484,8 +484,8 @@ echo_interval=n
sets the interval at which echo requests are sent to the server on an
idling connection. This setting also affects the time required for a
connection to an unresponsive server to timeout. Here n is the echo
- interval in seconds. The reconnection happens at twice the value of the
- echo_interval set for an unresponsive server.
+ interval in seconds. The reconnection happens at three times the
+ value of the echo_interval set for an unresponsive server.
If this option is not given then the default value of 60 seconds is used.
The minimum tunable value is 1 second and maximum can go up to 600 seconds.
--
2.53.0

View File

@ -1,49 +0,0 @@
From e7ec0032898d855be144c0cdc9d9e3f78ae01bf2 Mon Sep 17 00:00:00 2001
From: Paulo Alcantara <pc@manguebit.com>
Date: Sun, 10 Mar 2024 22:24:24 -0300
Subject: [PATCH 1/2] mount.cifs.rst: add missing reference for sssd
Reference sssd in mount.cifs(8) as it can be used instead of winbind
via cifs.idmap utility. It's also enabled by default in most systems.
Signed-off-by: Paulo Alcantara (Red Hat) <pc@manguebit.com>
---
mount.cifs.rst | 14 +++++++-------
1 file changed, 7 insertions(+), 7 deletions(-)
diff --git a/mount.cifs.rst b/mount.cifs.rst
index 3becf200e038..64127b23cf17 100644
--- a/mount.cifs.rst
+++ b/mount.cifs.rst
@@ -773,10 +773,10 @@ specified in the following Microsoft TechNet document:
In order to map SIDs to/from UIDs and GIDs, the following is required:
- a kernel upcall to the ``cifs.idmap`` utility set up via request-key.conf(5)
-- winbind support configured via nsswitch.conf(5) and smb.conf(5)
+- winbind or sssd support configured via nsswitch.conf(5)
-Please refer to the respective manpages of cifs.idmap(8) and
-winbindd(8) for more information.
+Please refer to the respective manpages of cifs.idmap(8), winbindd(8)
+and sssd(8) for more information.
Security descriptors for a file object can be retrieved and set
directly using extended attribute named ``system.cifs_acl``. The
@@ -792,10 +792,10 @@ Some of the things to consider while using this mount option:
- The mapping between a CIFS/NTFS ACL and POSIX file permission bits
is imperfect and some ACL information may be lost in the
translation.
-- If either upcall to cifs.idmap is not setup correctly or winbind is
- not configured and running, ID mapping will fail. In that case uid
- and gid will default to either to those values of the share or to
- the values of uid and/or gid mount options if specified.
+- If either upcall to cifs.idmap is not setup correctly or winbind or
+ sssd is not configured and running, ID mapping will fail. In that
+ case uid and gid will default to either to those values of the share
+ or to the values of uid and/or gid mount options if specified.
**********************************
ACCESSING FILES WITH BACKUP INTENT
--
2.44.0

View File

@ -1,59 +0,0 @@
From 4718e09e4b15b957bf9d729793bc3de7caad8134 Mon Sep 17 00:00:00 2001
From: Paulo Alcantara <pc@manguebit.com>
Date: Sun, 10 Mar 2024 22:24:25 -0300
Subject: [PATCH 2/2] mount.cifs.rst: update section about xattr/acl support
Update section about required xattr/acl support for UID/GID mapping.
Signed-off-by: Paulo Alcantara (Red Hat) <pc@manguebit.com>
---
mount.cifs.rst | 26 +++++++++++++++++++-------
1 file changed, 19 insertions(+), 7 deletions(-)
diff --git a/mount.cifs.rst b/mount.cifs.rst
index 64127b23cf17..d82a13c932b3 100644
--- a/mount.cifs.rst
+++ b/mount.cifs.rst
@@ -321,11 +321,12 @@ soft
noacl
Do not allow POSIX ACL operations even if server would support them.
- The CIFS client can get and set POSIX ACLs (getfacl, setfacl) to Samba
- servers version 3.0.10 and later. Setting POSIX ACLs requires enabling
- both ``CIFS_XATTR`` and then ``CIFS_POSIX`` support in the CIFS
- configuration options when building the cifs module. POSIX ACL support
- can be disabled on a per mount basis by specifying ``noacl`` on mount.
+ The CIFS client can get and set POSIX ACLs (getfacl, setfacl) to
+ Samba servers version 3.0.10 and later. Setting POSIX ACLs requires
+ enabling both ``CONFIG_CIFS_XATTR`` and then ``CONFIG_CIFS_POSIX``
+ support in the CIFS configuration options when building the cifs
+ module. POSIX ACL support can be disabled on a per mount basis by
+ specifying ``noacl`` on mount.
cifsacl
This option is used to map CIFS/NTFS ACLs to/from Linux permission
@@ -762,8 +763,19 @@ bits, and POSIX ACL as user authentication model. This is the most
common authentication model for CIFS servers and is the one used by
Windows.
-Support for this requires both CIFS_XATTR and CIFS_ACL support in the
-CIFS configuration options when building the cifs module.
+Support for this requires cifs kernel module built with both
+``CONFIG_CIFS_XATTR`` and ``CONFIG_CIFS_ACL`` options enabled. Since
+Linux 5.3, ``CONFIG_CIFS_ACL`` option no longer exists as CIFS/NTFS
+ACL support is always built into cifs kernel module.
+
+Most distribution kernels will already have those options enabled by
+default, but you can still check if they are enabled with::
+
+ cat /lib/modules/$(uname -r)/build/.config
+
+Alternatively, if kernel is configured with ``CONFIG_IKCONFIG_PROC``::
+
+ zcat /proc/config.gz
A CIFS/NTFS ACL is mapped to file permission bits using an algorithm
specified in the following Microsoft TechNet document:
--
2.44.0

View File

@ -3,7 +3,7 @@
Name: cifs-utils
Version: 7.0
Release: 5%{pre_release}%{?dist}
Release: 1%{pre_release}%{?dist}
Summary: Utilities for mounting and managing CIFS mounts
Group: System Environment/Daemons
@ -19,12 +19,6 @@ Requires(preun): /usr/sbin/alternatives
Source0: https://download.samba.org/pub/linux-cifs/cifs-utils/%{name}-%{version}.tar.bz2
Patch1: 0001-Use-explicit-usr-bin-python3.patch
Patch2: mount.cifs.rst-add-missing-reference-for-sssd.patch
Patch3: mount.cifs.rst-update-section-about-xattr-acl-suppor.patch
Patch4: docs-update-echo_interval-description.patch
Patch5: cifs.upcall-remove-getpwuid-dependency.patch
Patch6: cifs.upcall-fix-compiler-warning-with-Wvla.patch
Patch7: cifs.upcall-fix-regression-with-krb5-creduid.patch
%description
The SMB/CIFS protocol is a standard file sharing protocol widely deployed
@ -60,12 +54,6 @@ provide these credentials to the kernel automatically at login.
%prep
%setup -q -n %{name}-%{version}%{pre_release}
%patch1 -p1
%patch2 -p1
%patch3 -p1
%patch4 -p1
%patch5 -p1
%patch6 -p1
%patch7 -p1
%build
autoreconf -i
@ -125,23 +113,6 @@ fi
%{_mandir}/man8/pam_cifscreds.8.gz
%changelog
* Tue Jul 7 2026 Paulo Alcantara <paalcant@redhat.com> - 7.0-5
- resolves: RHEL-192933 - fix krb5 mount regression
* Thu Jun 25 2026 Paulo Alcantara <paalcant@redhat.com> - 7.0-4
- cifs.upcall: remove getpwuid dependency
- cifs.upcall: fix compiler warning with -Wvla
- Resolves: RHEL-185759 - Fix CVE-2026-12505
* Thu Feb 19 2026 Paulo Alcantara <paalcant@redhat.com> - 7.0-3
- docs: update echo_interval description
- Resolves: RHEL-80397
* Thu Jun 13 2024 Paulo Alcantara <paalcant@redhat.com> - 7.0-2
- mount.cifs.rst: add missing reference for sssd
- mount.cifs.rst: update section about xattr/acl support
- Resolves: RHEL-41059
* Mon Jan 30 2023 Pavel Filipenský <pfilipen@redhat.com> - 7.0-1
- Update to cifs-utils-7.0
- Resolves: rhbz#2163373