Mdmonitor: Fix segfault and improve logging
PATCH 1/1 Mdmonitor: Fix segfault and improve logging
From: Oleksandr Shchirskyi <[email protected]>
Check that devices passed to mdmonitor are md arrays. Also, change logging, so mdmonitor in verbose mode will report its configuration.
--- Monitor.c | 36 ++++++++++++++++++++++++------------ mdadm.h | 1 + mdopen.c | 17 +++++++++++++++++ util.c | 2 +- 4 files changed, 43 insertions(+), 13 deletions(-)
Monitor.c +24 -12
@@ -137,24 +137,27 @@ int Monitor(struct mddev_dev *devlist,struct mddev_ident *mdlist;int delay_for_event = c->delay; -if (!mailaddr) { +if (!mailaddr) mailaddr =conf_get_mailaddr (); -if (mailaddr && ! c->scan) -pr_err ("Monitor using email address \"%s\" from config file\n" , - mailaddr); - } - mailfrom =conf_get_mailfrom (); -if (!alert_cmd) { +if (!alert_cmd) alert_cmd =conf_get_program (); -if (alert_cmd && !c->scan) -pr_err ("Monitor using program \"%s\" from config file\n" , - alert_cmd); - } + + mailfrom =conf_get_mailfrom (); +if (c->scan && !mailaddr && !alert_cmd && !dosyslog) {pr_err ("No mail address or alert command - not monitoring.\n" );return 1 ; } + +if (c->verbose) { +pr_err ("Monitor is started with delay %ds\n" , c->delay); +if (mailaddr) +pr_err ("Monitor using email address %s\n" , mailaddr); +if (alert_cmd) +pr_err ("Monitor using program %s\n" , alert_cmd); + } + info.alert_cmd = alert_cmd; info.mailaddr = mailaddr; info.mailfrom = mailfrom;
@@ -183,6 +186,10 @@ int Monitor(struct mddev_dev *devlist,continue ;if (strcasecmp (mdlist->devname,"<ignore>" ) ==0 )continue ; + +if (!is_mddev (mdlist->devname)) +return 1 ; + st =xcalloc (1 ,sizeof *st);if (mdlist->devname[0 ] =='/' ) st->devname =xstrdup (mdlist->devname);
@@ -204,7 +211,12 @@ int Monitor(struct mddev_dev *devlist,struct mddev_dev *dv;for (dv = devlist; dv; dv = dv->next) { -struct state *st =xcalloc (1 ,sizeof *st); +struct state *st; + +if (!is_mddev (dv->devname)) +return 1 ; + + st =xcalloc (1 ,sizeof *st); mdlist =conf_get_ident (dv->devname); st->devname =xstrdup (dv->devname); st->next = statelist;
mdadm.h +1 -0
@@ -1607,6 +1607,7 @@ extern int create_mddev(char *dev, char *name, int autof, int trustworthy,#define FOREIGN 2#define METADATA 3extern int open_mddev (char *dev,int report_errors); +extern int is_mddev (char *dev);extern int open_container (int fd);extern int metadata_container_matches (char *metadata,char *devnm);extern int metadata_subdev_matches (char *metadata,char *devnm);
mdopen.c +17 -0
@@ -475,6 +475,23 @@ int open_mddev(char *dev, int report_errors)return mdfd; } +/** +* is_mddev() - check that file name passed is an md device. +* @dev: file name that has to be checked. +* Return: 1 if file passed is an md device, 0 if not. +*/ +int is_mddev (char *dev) +{ +int fd =open_mddev (dev,1 ); + +if (fd >=0 ) { +close (fd); +return 1 ; + } + +return 0 ; +} +char *find_free_devnm (int use_partitions) {static char devnm[32 ];
util.c +1 -1
@@ -268,7 +268,7 @@ int md_array_active(int fd)* GET_ARRAY_INFO doesn't provide access to the proper state * information, so fallback to a basic check for raid_disks != 0 */ - ret =ioctl (fd,GET_ARRAY_INFO , &array); + ret =md_get_array_info (fd, &array); }return !ret;
Signature · 1 line
2.26.2
Dear Kinga, dear Oleksandr,
Am 21.03.22 um 13:32 schrieb Kinga Tanska:
Quoted 5 lines
From: Oleksandr Shchirskyi <[email protected]>
Check that devices passed to mdmonitor are md arrays. Also, change logging, so mdmonitor in verbose mode will report its configuration.
Thank you for the patch. As these are two distinct changes, could you please split it into two commits?
If you could add a reproducer the segmentation fault, that’d also be great.
Kind regards,
Paul
Quoted 130 lines
Signed-off-by Oleksandr Shchirskyi <[email protected]>Signed-off-by Kinga Tanska <[email protected]>--- Monitor.c | 36 ++++++++++++++++++++++++------------ mdadm.h | 1 + mdopen.c | 17 +++++++++++++++++ util.c | 2 +- 4 files changed, 43 insertions(+), 13 deletions(-)Monitor.c +24
-12@@ -137,24 +137,27 @@ int Monitor(struct mddev_dev *devlist,struct mddev_ident *mdlist;int delay_for_event = c->delay; -if (!mailaddr) { +if (!mailaddr) mailaddr =conf_get_mailaddr (); -if (mailaddr && ! c->scan) -pr_err ("Monitor using email address \"%s\" from config file\n" , - mailaddr); - } - mailfrom =conf_get_mailfrom (); -if (!alert_cmd) { +if (!alert_cmd) alert_cmd =conf_get_program (); -if (alert_cmd && !c->scan) -pr_err ("Monitor using program \"%s\" from config file\n" , - alert_cmd); - } + + mailfrom =conf_get_mailfrom (); +if (c->scan && !mailaddr && !alert_cmd && !dosyslog) {pr_err ("No mail address or alert command - not monitoring.\n" );return 1 ; } + +if (c->verbose) { +pr_err ("Monitor is started with delay %ds\n" , c->delay); +if (mailaddr) +pr_err ("Monitor using email address %s\n" , mailaddr); +if (alert_cmd) +pr_err ("Monitor using program %s\n" , alert_cmd); + } + info.alert_cmd = alert_cmd; info.mailaddr = mailaddr; info.mailfrom = mailfrom;@@ -183,6 +186,10 @@ int Monitor(struct mddev_dev *devlist,continue ;if (strcasecmp (mdlist->devname,"<ignore>" ) ==0 )continue ; + +if (!is_mddev (mdlist->devname)) +return 1 ; + st =xcalloc (1 ,sizeof *st);if (mdlist->devname[0 ] =='/' ) st->devname =xstrdup (mdlist->devname);@@ -204,7 +211,12 @@ int Monitor(struct mddev_dev *devlist,struct mddev_dev *dv;for (dv = devlist; dv; dv = dv->next) { -struct state *st =xcalloc (1 ,sizeof *st); +struct state *st; + +if (!is_mddev (dv->devname)) +return 1 ; + + st =xcalloc (1 ,sizeof *st); mdlist =conf_get_ident (dv->devname); st->devname =xstrdup (dv->devname); st->next = statelist;mdadm.h +1
-0@@ -1607,6 +1607,7 @@ extern int create_mddev(char *dev, char *name, int autof, int trustworthy,#define FOREIGN 2#define METADATA 3extern int open_mddev (char *dev,int report_errors); +extern int is_mddev (char *dev);extern int open_container (int fd);extern int metadata_container_matches (char *metadata,char *devnm);extern int metadata_subdev_matches (char *metadata,char *devnm);mdopen.c +17
-0@@ -475,6 +475,23 @@ int open_mddev(char *dev, int report_errors)return mdfd; } +/** +* is_mddev() - check that file name passed is an md device. +* @dev: file name that has to be checked. +* Return: 1 if file passed is an md device, 0 if not. +*/ +int is_mddev (char *dev) +{ +int fd =open_mddev (dev,1 ); + +if (fd >=0 ) { +close (fd); +return 1 ; + } + +return 0 ; +} +char *find_free_devnm (int use_partitions) {static char devnm[32 ];util.c +1
-1@@ -268,7 +268,7 @@ int md_array_active(int fd)* GET_ARRAY_INFO doesn't provide access to the proper state * information, so fallback to a basic check for raid_disks != 0 */ - ret =ioctl (fd,GET_ARRAY_INFO , &array); + ret =md_get_array_info (fd, &array); }return !ret;
Hi,
I will send splitted patches in next messages. There are steps to reproduce:
1. Stop mdmonitor instance e.g.
# systemctl stop mdmonitor
2. Run mdmonitor for non-md device e.g.
# mdadm --monitor /dev/nvme1n1
Call trace:
#0 0x00007ffff7617518 in __strcpy_sse2_unaligned () from /usr/lib64/libc.so.6 #1 0x000000000042bc9e in check_array ()
#2 0x000000000042c7af in Monitor ()
#3 0x0000000000406edc in main ()
Regards,
Kinga Tanska
On 3/21/2022 1:45 PM, Paul Menzel wrote:
Dear Kinga, dear Oleksandr,
Am 21.03.22 um 13:32 schrieb Kinga Tanska:
Quoted 5 lines
From: Oleksandr Shchirskyi <[email protected]>
Check that devices passed to mdmonitor are md arrays. Also, change logging, so mdmonitor in verbose mode will report its configuration.
Thank you for the patch. As these are two distinct changes, could you please split it into two commits?
If you could add a reproducer the segmentation fault, that’d also be great.
Kind regards,
Paul
Quoted 127 lines
Signed-off-by Oleksandr Shchirskyi <[email protected]>Signed-off-by Kinga Tanska <[email protected]>---
Monitor.c| 36 ++++++++++++++++++++++++------------ mdadm.h | 1 +
mdopen.c | 17 +++++++++++++++++
util.c | 2 +-4 files changed, 43 insertions(+), 13 deletions(-)Monitor.c +14
-9@@ -137,24 +137,27 @@ int Monitor(struct mddev_dev *devlist, struct mddev_ident *mdlist; int delay_for_event = c->delay; - if (!mailaddr) { + if (!mailaddr) mailaddr = conf_get_mailaddr(); - if (mailaddr && ! c->scan) - pr_err("Monitor using email address \"%s\" from config file\n", - mailaddr); - } - mailfrom = conf_get_mailfrom(); - if (!alert_cmd) { + if (!alert_cmd) alert_cmd = conf_get_program(); - if (alert_cmd && !c->scan) - pr_err("Monitor using program \"%s\" from config file\n", - alert_cmd); - } + + mailfrom = conf_get_mailfrom(); + if (c->scan && !mailaddr && !alert_cmd && !dosyslog) { pr_err("No mail address or alert command - not monitoring.\n"); return 1; } + + if (c->verbose) { + pr_err("Monitor is started with delay %ds\n", c->delay); + if (mailaddr) + pr_err("Monitor using email address %s\n", mailaddr); + if (alert_cmd) + pr_err("Monitor using program %s\n", alert_cmd); + } + info.alert_cmd = alert_cmd; info.mailaddr = mailaddr;info.mailfrom = mailfrom;
Monitor.c +10
-1@@ -183,6 +186,10 @@ int Monitor(struct mddev_dev *devlist, continue; if (strcasecmp(mdlist->devname, "<ignore>") == 0) continue; + + if (!is_mddev(mdlist->devname)) + return 1; + st = xcalloc(1, sizeof *st); if (mdlist->devname[0] == '/') st->devname = xstrdup(mdlist->devname);@@ -204,7 +211,12 @@ int Monitor(struct mddev_dev *devlist, struct mddev_dev *dv; for (dv = devlist; dv; dv = dv->next) { - struct state *st = xcalloc(1, sizeof *st); + struct state *st; + + if (!is_mddev(dv->devname)) + return 1; + + st = xcalloc(1, sizeof *st); mdlist = conf_get_ident(dv->devname); st->devname = xstrdup(dv->devname); st->next = statelist;mdadm.h +1
-0@@ -1607,6 +1607,7 @@ extern int create_mddev(char *dev, char *name, int autof, int trustworthy, #define FOREIGN 2 #define METADATA 3 extern int open_mddev(char *dev, int report_errors); +extern int is_mddev (char *dev); extern int open_container(int fd); extern int metadata_container_matches(char *metadata, char *devnm);extern int
metadata_subdev_matches(char *metadata, char *devnm);mdopen.c +0
-0@@ -475,6 +475,23 @@ int open_mddev(char *dev, int report_errors) return mdfd; }+/**
+ *is_mddev()- check that file name passed is an md device.
+ * @dev: file name that has to be checked.
+ * Return: 1 if file passed is an md device, 0 if not.
+ */mdopen.c +12
-0+int is_mddev (char *dev) +{ + int fd = open_mddev(dev, 1); + + if (fd >= 0) { + close(fd); + return 1; + } + + return 0; +} +char *find_free_devnm(int
use_partitions)
{
static char devnm[32];util.c +1
-1@@ -268,7 +268,7 @@ int md_array_active(int fd)* GET_ARRAY_INFO doesn't provide access to the proper state * information, so fallback to a basic check for raid_disks != 0 */ - ret = ioctl(fd, GET_ARRAY_INFO, &array); + ret = md_get_array_info(fd, &array); } return !ret;
Dear Kinga,
Am 23.03.22 um 10:02 schrieb Tanska, Kinga:
Quoted 1 line
I will send splitted patches in next messages.
Thank you.
Quoted 11 lines
There are steps to reproduce:
1. Stop mdmonitor instance e.g.
# systemctl stop mdmonitor2. Run mdmonitor for non-md device e.g.
# mdadm --monitor /dev/nvme1n1
Quoted 6 lines
Call trace:#0
0x00007ffff7617518in__strcpy_sse2_unaligned() from/usr/lib64/libc.so.6#10x000000000042bc9eincheck_array()
#20x000000000042c7afin Monitor ()
#30x0000000000406edcin main ()
I am unable to reproduce it with mdadm 4.1:
$ sudo strace mdadm --monitor /dev/sda
execve("/usr/bin/mdadm", ["mdadm", "--monitor", "/dev/sda"],
0x7ffc8b00e880 /* 14 vars */) = 0
[…]
openat(AT_FDCWD, "/dev/sda", O_RDONLY) = 3
fstat(3, {st_mode=S_IFBLK|0660, st_rdev=makedev(0x8, 0), ...}) = 0
readlink("/sys/dev/block/8:0", "../../devices/pci0000:00/0000:00"..., 199) = 91
close(3) = 0
pselect6(1, NULL, NULL, [], {tv_sec=1000, tv_nsec=0}, NULL) = ?
ERESTARTNOHAND (To be restarted if no handler)
--- SIGWINCH {si_signo=SIGWINCH, si_code=SI_KERNEL} ---
pselect6(1, NULL, NULL, [], {tv_sec=999, tv_nsec=568020697}, NULL)
= ? ERESTARTNOHAND (To be restarted if no handler)
--- SIGWINCH {si_signo=SIGWINCH, si_code=SI_KERNEL} ---
pselect6(1, NULL, NULL, [], {tv_sec=999, tv_nsec=521836826}, NULLDid a specific commit introduce the problem? If so, it’d be great if you added a Fixes line.
[…]
Kind regards,
Paul
Hello Paul,
I've tried to reproduce it with mdadm 4.1 and mdadm 4.2 and both apps exit with
segmentation fault. This patch is not a fix, it adds checking if device is md array.
It hasn't been checked before in mdmonitor, so segmentation fault should appear.
Regards,
Kinga
On 3/23/2022 11:31 AM, Paul Menzel wrote:
Dear Kinga,
Am 23.03.22 um 10:02 schrieb Tanska, Kinga:
Quoted 1 line
I will send splitted patches in next messages.
Thank you.
Quoted 11 lines
There are steps to reproduce:
1. Stop mdmonitor instance e.g.
# systemctl stop mdmonitor2. Run mdmonitor for non-md device e.g.
# mdadm --monitor /dev/nvme1n1
Quoted 6 lines
Call trace:#0
0x00007ffff7617518in__strcpy_sse2_unaligned() from/usr/lib64/libc.so.6#10x000000000042bc9eincheck_array()
#20x000000000042c7afin Monitor ()
#30x0000000000406edcin main ()
I am unable to reproduce it with mdadm 4.1:
$ sudo strace mdadm --monitor /dev/sda
execve("/usr/bin/mdadm", ["mdadm", "--monitor", "/dev/sda"], 0x7ffc8b00e880 /* 14 vars */) = 0
[…]
openat(AT_FDCWD, "/dev/sda", O_RDONLY) = 3
fstat(3, {st_mode=S_IFBLK|0660, st_rdev=makedev(0x8, 0), ...}) = 0
readlink("/sys/dev/block/8:0", "../../devices/pci0000:00/0000:00"..., 199) = 91
close(3) = 0
pselect6(1, NULL, NULL, [], {tv_sec=1000, tv_nsec=0}, NULL) = ? ERESTARTNOHAND (To be restarted if no handler)
--- SIGWINCH {si_signo=SIGWINCH, si_code=SI_KERNEL} ---
pselect6(1, NULL, NULL, [], {tv_sec=999, tv_nsec=568020697}, NULL) = ? ERESTARTNOHAND (To be restarted if no handler)
--- SIGWINCH {si_signo=SIGWINCH, si_code=SI_KERNEL} ---
pselect6(1, NULL, NULL, [], {tv_sec=999, tv_nsec=521836826}, NULLDid a specific commit introduce the problem? If so, it’d be great if you added a Fixes line.
[…]
Kind regards,
Paul
Dear Kinga,
Am 04.04.22 um 14:39 schrieb Tanska, Kinga:
Quoted 6 lines
I've tried to reproduce it with mdadm 4.1 and mdadm 4.2 and both apps
exit with segmentation fault. This patch is not a fix, it adds
checking if device is md array.It hasn't been checked before in mdmonitor, so segmentation fault should appear.
Understood, still it’d be nice to have a way to reproduce this. So please elaborate in the commit message, and give more details about your test setup.
Kind regards,
Paul