| Message ID | 538F84DF.3090300@suse.com (mailing list archive) |
|---|---|
| State | Accepted |
| Headers | show |
On Wed, Jun 04, 2014 at 04:43:11PM -0400, Jeff Mahoney wrote: > --- a/utils.c > +++ b/utils.c > @@ -987,6 +987,63 @@ static int blk_file_in_dev_list(struct b > } > > /* > + * Resolve a pathname to a device mapper node to /dev/mapper/<name> > + * Returns NULL on invalid input or malloc failure; Other failures > + * will be handled by the caller using the input pathame. > + */ > +char *canonicalize_dm_name(const char *ptname) > +{ > + FILE *f; > + size_t sz; > + char path[256], name[256], *res = NULL; FYI, I've changed this to PATH_MAX -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Jeff, Nice patch. However its better if we do this in the btrfs kernel function btrfs_scan_one_device(). Since the non-canonicalize path can still sneak through the btrfs specific mount option "device=". Any comments ? Thanks, Anand On 06/05/2014 04:43 AM, Jeff Mahoney wrote: > mount(8) will canonicalize pathnames before passing them to the kernel. > Links to e.g. /dev/sda will be resolved to /dev/sda. Links to /dev/dm-# > will be resolved using the name of the device mapper table to > /dev/mapper/<name>. > > Btrfs will use whatever name the user passes to it, regardless of whether > it is canonical or not. That means that if a 'btrfs device ready' is > issued on any device node pointing to the original device, it will adopt > the new name instead of the name that was used during mount. > > Mounting using /dev/sdb2 will result in df: > /dev/sdb2 209715200 39328 207577088 1% /mnt > > # ls -la /dev/whatever-i-like > lrwxrwxrwx 1 root root 4 Jun 4 13:36 /dev/whatever-i-like -> sdb2 > # btrfs dev ready /dev/whatever-i-like > # df /mnt > /dev/whatever-i-like 209715200 39328 207577088 1% /mnt > > Likewise, mounting with /dev/mapper/whatever and using /dev/dm-0 with a > btrfs device command results in df showing /dev/dm-0. This can happen with > multipath devices with friendly names enabled and doing something like > 'partprobe' which (at least with our version) ends up issuing a 'change' > uevent on the sysfs node. That *always* uses the dm-# name, and we get > confused users. > > This patch does the same canonicalization of the paths that mount does > so that we don't end up having inconsistent names reported by ->show_devices > later. > > Signed-off-by: Jeff Mahoney <jeffm@suse.com> > --- > cmds-device.c | 60 ++++++++++++++++++++++++++++++++++++++++++++------------- > cmds-replace.c | 13 ++++++++++-- > utils.c | 57 ++++++++++++++++++++++++++++++++++++++++++++++++++++++ > utils.h | 2 + > 4 files changed, 117 insertions(+), 15 deletions(-) > > --- a/cmds-device.c > +++ b/cmds-device.c > @@ -95,6 +95,7 @@ static int cmd_add_dev(int argc, char ** > int devfd, res; > u64 dev_block_count = 0; > int mixed = 0; > + char *path; > > res = test_dev_for_mkfs(argv[i], force, estr); > if (res) { > @@ -118,15 +119,24 @@ static int cmd_add_dev(int argc, char ** > goto error_out; > } > > - strncpy_null(ioctl_args.name, argv[i]); > + path = canonicalize_path(argv[i]); > + if (!path) { > + fprintf(stderr, > + "ERROR: Could not canonicalize pathname '%s': %s\n", > + argv[i], strerror(errno)); > + ret++; > + goto error_out; > + } > + > + strncpy_null(ioctl_args.name, path); > res = ioctl(fdmnt, BTRFS_IOC_ADD_DEV, &ioctl_args); > e = errno; > - if(res<0){ > + if (res < 0) { > fprintf(stderr, "ERROR: error adding the device '%s' - %s\n", > - argv[i], strerror(e)); > + path, strerror(e)); > ret++; > } > - > + free(path); > } > > error_out: > @@ -242,6 +252,7 @@ static int cmd_scan_dev(int argc, char * > > for( i = devstart ; i < argc ; i++ ){ > struct btrfs_ioctl_vol_args args; > + char *path; > > if (!is_block_device(argv[i])) { > fprintf(stderr, > @@ -249,9 +260,17 @@ static int cmd_scan_dev(int argc, char * > ret = 1; > goto close_out; > } > - printf("Scanning for Btrfs filesystems in '%s'\n", argv[i]); > + path = canonicalize_path(argv[i]); > + if (!path) { > + fprintf(stderr, > + "ERROR: Could not canonicalize path '%s': %s\n", > + argv[i], strerror(errno)); > + ret = 1; > + goto close_out; > + } > + printf("Scanning for Btrfs filesystems in '%s'\n", path); > > - strncpy_null(args.name, argv[i]); > + strncpy_null(args.name, path); > /* > * FIXME: which are the error code returned by this ioctl ? > * it seems that is impossible to understand if there no is > @@ -262,9 +281,11 @@ static int cmd_scan_dev(int argc, char * > > if( ret < 0 ){ > fprintf(stderr, "ERROR: unable to scan the device '%s' - %s\n", > - argv[i], strerror(e)); > + path, strerror(e)); > + free(path); > goto close_out; > } > + free(path); > } > > close_out: > @@ -284,6 +305,7 @@ static int cmd_ready_dev(int argc, char > struct btrfs_ioctl_vol_args args; > int fd; > int ret; > + char *path; > > if (check_argc_min(argc, 2)) > usage(cmd_ready_dev_usage); > @@ -293,22 +315,34 @@ static int cmd_ready_dev(int argc, char > perror("failed to open /dev/btrfs-control"); > return 1; > } > - if (!is_block_device(argv[1])) { > + > + path = canonicalize_path(argv[argc - 1]); > + if (!path) { > fprintf(stderr, > - "ERROR: %s is not a block device\n", argv[1]); > - close(fd); > - return 1; > + "ERROR: Could not canonicalize pathname '%s': %s\n", > + argv[argc - 1], strerror(errno)); > + ret = 1; > + goto out; > } > > - strncpy(args.name, argv[argc - 1], BTRFS_PATH_NAME_MAX); > + if (!is_block_device(path)) { > + fprintf(stderr, > + "ERROR: %s is not a block device\n", path); > + ret = 1; > + goto out; > + } > + > + strncpy(args.name, path, BTRFS_PATH_NAME_MAX); > ret = ioctl(fd, BTRFS_IOC_DEVICES_READY, &args); > if (ret < 0) { > fprintf(stderr, "ERROR: unable to determine if the device '%s'" > - " is ready for mounting - %s\n", argv[argc - 1], > + " is ready for mounting - %s\n", path, > strerror(errno)); > ret = 1; > } > > +out: > + free(path); > close(fd); > return ret; > } > --- a/cmds-replace.c > +++ b/cmds-replace.c > @@ -134,7 +134,7 @@ static int cmd_start_replace(int argc, c > int fddstdev = -1; > char *path; > char *srcdev; > - char *dstdev; > + char *dstdev = NULL; > int avoid_reading_from_srcdev = 0; > int force_using_targetdev = 0; > struct stat st; > @@ -204,7 +204,12 @@ static int cmd_start_replace(int argc, c > } > > srcdev = argv[optind]; > - dstdev = argv[optind + 1]; > + dstdev = canonicalize_path(argv[optind + 1]); > + if (!dstdev) { > + fprintf(stderr, > + "ERROR: Could not canonicalize path '%s': %s\n", > + argv[optind + 1], strerror(errno)); > + } > > if (is_numerical(srcdev)) { > struct btrfs_ioctl_fs_info_args fi_args; > @@ -278,6 +283,8 @@ static int cmd_start_replace(int argc, c > > close(fddstdev); > fddstdev = -1; > + free(dstdev); > + dstdev = NULL; > > dev_replace_handle_sigint(fdmnt); > if (!do_not_background) { > @@ -312,6 +319,8 @@ static int cmd_start_replace(int argc, c > return 0; > > leave_with_error: > + if (dstdev) > + free(dstdev); > if (fdmnt != -1) > close(fdmnt); > if (fdsrcdev != -1) > --- a/utils.c > +++ b/utils.c > @@ -987,6 +987,63 @@ static int blk_file_in_dev_list(struct b > } > > /* > + * Resolve a pathname to a device mapper node to /dev/mapper/<name> > + * Returns NULL on invalid input or malloc failure; Other failures > + * will be handled by the caller using the input pathame. > + */ > +char *canonicalize_dm_name(const char *ptname) > +{ > + FILE *f; > + size_t sz; > + char path[256], name[256], *res = NULL; > + > + if (!ptname || !*ptname) > + return NULL; > + > + snprintf(path, sizeof(path), "/sys/block/%s/dm/name", ptname); > + if (!(f = fopen(path, "r"))) > + return NULL; > + > + /* read <name>\n from sysfs */ > + if (fgets(name, sizeof(name), f) && (sz = strlen(name)) > 1) { > + name[sz - 1] = '\0'; > + snprintf(path, sizeof(path), "/dev/mapper/%s", name); > + > + if (access(path, F_OK) == 0) > + res = strdup(path); > + } > + fclose(f); > + return res; > +} > + > +/* > + * Resolve a pathname to a canonical device node, e.g. /dev/sda1 or > + * to a device mapper pathname. > + * Returns NULL on invalid input or malloc failure; Other failures > + * will be handled by the caller using the input pathame. > + */ > +char *canonicalize_path(const char *path) > +{ > + char *canonical, *p; > + > + if (!path || !*path) > + return NULL; > + > + canonical = realpath(path, NULL); > + if (!canonical) > + return strdup(path); > + p = strrchr(canonical, '/'); > + if (p && strncmp(p, "/dm-", 4) == 0 && isdigit(*(p + 4))) { > + char *dm = canonicalize_dm_name(p + 1); > + if (dm) { > + free(canonical); > + return dm; > + } > + } > + return canonical; > +} > + > +/* > * returns 1 if the device was mounted, < 0 on error or 0 if everything > * is safe to continue. > */ > --- a/utils.h > +++ b/utils.h > @@ -61,6 +61,8 @@ int btrfs_add_to_fsid(struct btrfs_trans > int btrfs_scan_for_fsid(int run_ioctls); > void btrfs_register_one_device(char *fname); > int btrfs_scan_one_dir(char *dirname, int run_ioctl); > +char *canonicalize_dm_name(const char *ptname); > +char *canonicalize_path(const char *path); > int check_mounted(const char *devicename); > int check_mounted_where(int fd, const char *file, char *where, int size, > struct btrfs_fs_devices **fs_devices_mnt); > -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On 9/13/14, 3:31 AM, Anand Jain wrote: > > > Jeff, > > Nice patch. However its better if we do this in the btrfs kernel > function btrfs_scan_one_device(). Since the non-canonicalize path > can still sneak through the btrfs specific mount option "device=". > > Any comments ? My initial reaction is to avoid playing naming names within the kernel. But since it's device mapper-specific, it might not be too messy to do that. In addition to the patch to the progs, we're also carrying a patch to systemd since it has it's own little ioctl wrapper to do the scanning. It needed to be fixed there as well. - -Jeff > Thanks, Anand > > > > On 06/05/2014 04:43 AM, Jeff Mahoney wrote: >> mount(8) will canonicalize pathnames before passing them to the >> kernel. Links to e.g. /dev/sda will be resolved to /dev/sda. >> Links to /dev/dm-# will be resolved using the name of the device >> mapper table to /dev/mapper/<name>. >> >> Btrfs will use whatever name the user passes to it, regardless of >> whether it is canonical or not. That means that if a 'btrfs >> device ready' is issued on any device node pointing to the >> original device, it will adopt the new name instead of the name >> that was used during mount. >> >> Mounting using /dev/sdb2 will result in df: /dev/sdb2 >> 209715200 39328 207577088 1% /mnt >> >> # ls -la /dev/whatever-i-like lrwxrwxrwx 1 root root 4 Jun 4 >> 13:36 /dev/whatever-i-like -> sdb2 # btrfs dev ready >> /dev/whatever-i-like # df /mnt /dev/whatever-i-like 209715200 >> 39328 207577088 1% /mnt >> >> Likewise, mounting with /dev/mapper/whatever and using /dev/dm-0 >> with a btrfs device command results in df showing /dev/dm-0. This >> can happen with multipath devices with friendly names enabled and >> doing something like 'partprobe' which (at least with our >> version) ends up issuing a 'change' uevent on the sysfs node. >> That *always* uses the dm-# name, and we get confused users. >> >> This patch does the same canonicalization of the paths that mount >> does so that we don't end up having inconsistent names reported >> by ->show_devices later. >> >> Signed-off-by: Jeff Mahoney <jeffm@suse.com> --- cmds-device.c | >> 60 ++++++++++++++++++++++++++++++++++++++++++++------------- >> cmds-replace.c | 13 ++++++++++-- utils.c | 57 >> ++++++++++++++++++++++++++++++++++++++++++++++++++++++ utils.h >> | 2 + 4 files changed, 117 insertions(+), 15 deletions(-) >> >> --- a/cmds-device.c +++ b/cmds-device.c @@ -95,6 +95,7 @@ static >> int cmd_add_dev(int argc, char ** int devfd, res; u64 >> dev_block_count = 0; int mixed = 0; + char *path; >> >> res = test_dev_for_mkfs(argv[i], force, estr); if (res) { @@ >> -118,15 +119,24 @@ static int cmd_add_dev(int argc, char ** goto >> error_out; } >> >> - strncpy_null(ioctl_args.name, argv[i]); + path = >> canonicalize_path(argv[i]); + if (!path) { + >> fprintf(stderr, + "ERROR: Could not canonicalize >> pathname '%s': %s\n", + argv[i], >> strerror(errno)); + ret++; + goto >> error_out; + } + + strncpy_null(ioctl_args.name, >> path); res = ioctl(fdmnt, BTRFS_IOC_ADD_DEV, &ioctl_args); e = >> errno; - if(res<0){ + if (res < 0) { >> fprintf(stderr, "ERROR: error adding the device '%s' - %s\n", - >> argv[i], strerror(e)); + path, strerror(e)); >> ret++; } - + free(path); } >> >> error_out: @@ -242,6 +252,7 @@ static int cmd_scan_dev(int argc, >> char * >> >> for( i = devstart ; i < argc ; i++ ){ struct btrfs_ioctl_vol_args >> args; + char *path; >> >> if (!is_block_device(argv[i])) { fprintf(stderr, @@ -249,9 >> +260,17 @@ static int cmd_scan_dev(int argc, char * ret = 1; goto >> close_out; } - printf("Scanning for Btrfs filesystems in >> '%s'\n", argv[i]); + path = canonicalize_path(argv[i]); + >> if (!path) { + fprintf(stderr, + >> "ERROR: Could not canonicalize path '%s': %s\n", + >> argv[i], strerror(errno)); + ret = 1; + >> goto close_out; + } + printf("Scanning for Btrfs >> filesystems in '%s'\n", path); >> >> - strncpy_null(args.name, argv[i]); + >> strncpy_null(args.name, path); /* * FIXME: which are the error >> code returned by this ioctl ? * it seems that is impossible to >> understand if there no is @@ -262,9 +281,11 @@ static int >> cmd_scan_dev(int argc, char * >> >> if( ret < 0 ){ fprintf(stderr, "ERROR: unable to scan the device >> '%s' - %s\n", - argv[i], strerror(e)); + >> path, strerror(e)); + free(path); goto close_out; } + >> free(path); } >> >> close_out: @@ -284,6 +305,7 @@ static int cmd_ready_dev(int argc, >> char struct btrfs_ioctl_vol_args args; int fd; int ret; >> + char *path; >> >> if (check_argc_min(argc, 2)) usage(cmd_ready_dev_usage); @@ >> -293,22 +315,34 @@ static int cmd_ready_dev(int argc, char >> perror("failed to open /dev/btrfs-control"); return 1; } - if >> (!is_block_device(argv[1])) { + + path = >> canonicalize_path(argv[argc - 1]); + if (!path) { >> fprintf(stderr, - "ERROR: %s is not a block device\n", >> argv[1]); - close(fd); - return 1; + >> "ERROR: Could not canonicalize pathname '%s': %s\n", + >> argv[argc - 1], strerror(errno)); + ret = 1; + goto >> out; } >> >> - strncpy(args.name, argv[argc - 1], BTRFS_PATH_NAME_MAX); + >> if (!is_block_device(path)) { + fprintf(stderr, + >> "ERROR: %s is not a block device\n", path); + ret = 1; + >> goto out; + } + + strncpy(args.name, path, >> BTRFS_PATH_NAME_MAX); ret = ioctl(fd, BTRFS_IOC_DEVICES_READY, >> &args); if (ret < 0) { fprintf(stderr, "ERROR: unable to >> determine if the device '%s'" - " is ready for >> mounting - %s\n", argv[argc - 1], + " is ready for >> mounting - %s\n", path, strerror(errno)); ret = 1; } >> >> +out: + free(path); close(fd); return ret; } --- >> a/cmds-replace.c +++ b/cmds-replace.c @@ -134,7 +134,7 @@ static >> int cmd_start_replace(int argc, c int fddstdev = -1; char *path; >> char *srcdev; - char *dstdev; + char *dstdev = NULL; int >> avoid_reading_from_srcdev = 0; int force_using_targetdev = 0; >> struct stat st; @@ -204,7 +204,12 @@ static int >> cmd_start_replace(int argc, c } >> >> srcdev = argv[optind]; - dstdev = argv[optind + 1]; + >> dstdev = canonicalize_path(argv[optind + 1]); + if (!dstdev) >> { + fprintf(stderr, + "ERROR: Could not >> canonicalize path '%s': %s\n", + argv[optind + 1], >> strerror(errno)); + } >> >> if (is_numerical(srcdev)) { struct btrfs_ioctl_fs_info_args >> fi_args; @@ -278,6 +283,8 @@ static int cmd_start_replace(int >> argc, c >> >> close(fddstdev); fddstdev = -1; + free(dstdev); + dstdev = >> NULL; >> >> dev_replace_handle_sigint(fdmnt); if (!do_not_background) { @@ >> -312,6 +319,8 @@ static int cmd_start_replace(int argc, c return >> 0; >> >> leave_with_error: + if (dstdev) + free(dstdev); if >> (fdmnt != -1) close(fdmnt); if (fdsrcdev != -1) --- a/utils.c +++ >> b/utils.c @@ -987,6 +987,63 @@ static int >> blk_file_in_dev_list(struct b } >> >> /* + * Resolve a pathname to a device mapper node to >> /dev/mapper/<name> + * Returns NULL on invalid input or malloc >> failure; Other failures + * will be handled by the caller using >> the input pathame. + */ +char *canonicalize_dm_name(const char >> *ptname) +{ + FILE *f; + size_t sz; + char >> path[256], name[256], *res = NULL; + + if (!ptname || >> !*ptname) + return NULL; + + snprintf(path, >> sizeof(path), "/sys/block/%s/dm/name", ptname); + if (!(f = >> fopen(path, "r"))) + return NULL; + + /* read <name>\n >> from sysfs */ + if (fgets(name, sizeof(name), f) && (sz = >> strlen(name)) > 1) { + name[sz - 1] = '\0'; + >> snprintf(path, sizeof(path), "/dev/mapper/%s", name); + + >> if (access(path, F_OK) == 0) + res = strdup(path); + >> } + fclose(f); + return res; +} + +/* + * Resolve a >> pathname to a canonical device node, e.g. /dev/sda1 or + * to a >> device mapper pathname. + * Returns NULL on invalid input or >> malloc failure; Other failures + * will be handled by the caller >> using the input pathame. + */ +char *canonicalize_path(const char >> *path) +{ + char *canonical, *p; + + if (!path || !*path) + >> return NULL; + + canonical = realpath(path, NULL); + if >> (!canonical) + return strdup(path); + p = >> strrchr(canonical, '/'); + if (p && strncmp(p, "/dm-", 4) == 0 >> && isdigit(*(p + 4))) { + char *dm = >> canonicalize_dm_name(p + 1); + if (dm) { + >> free(canonical); + return dm; + } + } + >> return canonical; +} + +/* * returns 1 if the device was mounted, >> < 0 on error or 0 if everything * is safe to continue. */ --- >> a/utils.h +++ b/utils.h @@ -61,6 +61,8 @@ int >> btrfs_add_to_fsid(struct btrfs_trans int btrfs_scan_for_fsid(int >> run_ioctls); void btrfs_register_one_device(char *fname); int >> btrfs_scan_one_dir(char *dirname, int run_ioctl); +char >> *canonicalize_dm_name(const char *ptname); +char >> *canonicalize_path(const char *path); int check_mounted(const >> char *devicename); int check_mounted_where(int fd, const char >> *file, char *where, int size, struct btrfs_fs_devices >> **fs_devices_mnt); >> > -- To unsubscribe from this list: send the line "unsubscribe > linux-btrfs" in the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html > - -- Jeff Mahoney SUSE Labs -----BEGIN PGP SIGNATURE----- Version: GnuPG/MacGPG2 v2.0.22 (Darwin) iQIcBAEBAgAGBQJUFhr9AAoJEB57S2MheeWyXzgQAMKR0b29uU0kyYbalOxjpVkB yAvR5kFItcEgRSfUUyqJqL+atJZ2PrzU4fQsz6i25cOBfQC5EdZYO/QENXe7HleU oCECYucZC14qTLwswhpMeCtyMrX98sjhQxZ6NPSmLZWi/Btc2QDdwwbij4B3KRnQ lE/AWmtT/J0fqlmjr18ETdTjTCZVyxUpN809hZpHAwQAiDZhXqbTUveaFu3DpTyn vDdM5LZ2oGPjS/3WgoNiFy36R1cnxIh3k/+aM/afU20T//dOn8Cdrh+k868hv9fV f3YfkyoHzwBCXp8VaOZytkOy7HCLbTEpnnB7MbvrNo1ukOlrFZRVRMSUrco2QD6x 4UYfgBj/zuzxIUZotEnpAmE8ppbMBylqjx+dyIwviDCU0Huw9mEnraBf3Yunuf3k za92pv/SYJnEFRl2w5ULhtR6GJ3RM8TQJOtWRJjPLkS3L30xvVOQZ1zXoazi/4yX qp7bKKN/hzC8mPvp8aoCW5c+F652e4pmUplbR4RDaNCD7ipv60XfQBVYOJOGh2YX +OCYERzFs4th3uPlKNDSfsLHwGY36bPC3mPdiaJW144MX1vHL5YKJiEKcnk5QZNY BwWidSvdylitAmZeKeCRTOJJh/b+Q7/K72RdWSl6g5qipa/00Gc97Sl/jNOyZ7K6 URPW75jcC/KtA2gIArtL =fAyc -----END PGP SIGNATURE----- -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 15/09/2014 06:47, Jeff Mahoney wrote: > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > On 9/13/14, 3:31 AM, Anand Jain wrote: >> >> >> Jeff, >> >> Nice patch. However its better if we do this in the btrfs kernel >> function btrfs_scan_one_device(). Since the non-canonicalize path >> can still sneak through the btrfs specific mount option "device=". >> >> Any comments ? > > My initial reaction is to avoid playing naming names within the > kernel. But since it's device mapper-specific, it might not be too > messy to do that. In addition to the patch to the progs, we're also > carrying a patch to systemd since it has it's own little ioctl wrapper > to do the scanning. It needed to be fixed there as well. looks like we need to fix systemd as well. As I check systemd is using READY ioctl. Thanks, Anand > - -Jeff > > >> Thanks, Anand >> >> >> >> On 06/05/2014 04:43 AM, Jeff Mahoney wrote: >>> mount(8) will canonicalize pathnames before passing them to the >>> kernel. Links to e.g. /dev/sda will be resolved to /dev/sda. >>> Links to /dev/dm-# will be resolved using the name of the device >>> mapper table to /dev/mapper/<name>. >>> >>> Btrfs will use whatever name the user passes to it, regardless of >>> whether it is canonical or not. That means that if a 'btrfs >>> device ready' is issued on any device node pointing to the >>> original device, it will adopt the new name instead of the name >>> that was used during mount. >>> >>> Mounting using /dev/sdb2 will result in df: /dev/sdb2 >>> 209715200 39328 207577088 1% /mnt >>> >>> # ls -la /dev/whatever-i-like lrwxrwxrwx 1 root root 4 Jun 4 >>> 13:36 /dev/whatever-i-like -> sdb2 # btrfs dev ready >>> /dev/whatever-i-like # df /mnt /dev/whatever-i-like 209715200 >>> 39328 207577088 1% /mnt >>> >>> Likewise, mounting with /dev/mapper/whatever and using /dev/dm-0 >>> with a btrfs device command results in df showing /dev/dm-0. This >>> can happen with multipath devices with friendly names enabled and >>> doing something like 'partprobe' which (at least with our >>> version) ends up issuing a 'change' uevent on the sysfs node. >>> That *always* uses the dm-# name, and we get confused users. >>> >>> This patch does the same canonicalization of the paths that mount >>> does so that we don't end up having inconsistent names reported >>> by ->show_devices later. >>> >>> Signed-off-by: Jeff Mahoney <jeffm@suse.com> --- cmds-device.c | >>> 60 ++++++++++++++++++++++++++++++++++++++++++++------------- >>> cmds-replace.c | 13 ++++++++++-- utils.c | 57 >>> ++++++++++++++++++++++++++++++++++++++++++++++++++++++ utils.h >>> | 2 + 4 files changed, 117 insertions(+), 15 deletions(-) >>> >>> --- a/cmds-device.c +++ b/cmds-device.c @@ -95,6 +95,7 @@ static >>> int cmd_add_dev(int argc, char ** int devfd, res; u64 >>> dev_block_count = 0; int mixed = 0; + char *path; >>> >>> res = test_dev_for_mkfs(argv[i], force, estr); if (res) { @@ >>> -118,15 +119,24 @@ static int cmd_add_dev(int argc, char ** goto >>> error_out; } >>> >>> - strncpy_null(ioctl_args.name, argv[i]); + path = >>> canonicalize_path(argv[i]); + if (!path) { + >>> fprintf(stderr, + "ERROR: Could not canonicalize >>> pathname '%s': %s\n", + argv[i], >>> strerror(errno)); + ret++; + goto >>> error_out; + } + + strncpy_null(ioctl_args.name, >>> path); res = ioctl(fdmnt, BTRFS_IOC_ADD_DEV, &ioctl_args); e = >>> errno; - if(res<0){ + if (res < 0) { >>> fprintf(stderr, "ERROR: error adding the device '%s' - %s\n", - >>> argv[i], strerror(e)); + path, strerror(e)); >>> ret++; } - + free(path); } >>> >>> error_out: @@ -242,6 +252,7 @@ static int cmd_scan_dev(int argc, >>> char * >>> >>> for( i = devstart ; i < argc ; i++ ){ struct btrfs_ioctl_vol_args >>> args; + char *path; >>> >>> if (!is_block_device(argv[i])) { fprintf(stderr, @@ -249,9 >>> +260,17 @@ static int cmd_scan_dev(int argc, char * ret = 1; goto >>> close_out; } - printf("Scanning for Btrfs filesystems in >>> '%s'\n", argv[i]); + path = canonicalize_path(argv[i]); + >>> if (!path) { + fprintf(stderr, + >>> "ERROR: Could not canonicalize path '%s': %s\n", + >>> argv[i], strerror(errno)); + ret = 1; + >>> goto close_out; + } + printf("Scanning for Btrfs >>> filesystems in '%s'\n", path); >>> >>> - strncpy_null(args.name, argv[i]); + >>> strncpy_null(args.name, path); /* * FIXME: which are the error >>> code returned by this ioctl ? * it seems that is impossible to >>> understand if there no is @@ -262,9 +281,11 @@ static int >>> cmd_scan_dev(int argc, char * >>> >>> if( ret < 0 ){ fprintf(stderr, "ERROR: unable to scan the device >>> '%s' - %s\n", - argv[i], strerror(e)); + >>> path, strerror(e)); + free(path); goto close_out; } + >>> free(path); } >>> >>> close_out: @@ -284,6 +305,7 @@ static int cmd_ready_dev(int argc, >>> char struct btrfs_ioctl_vol_args args; int fd; int ret; >>> + char *path; >>> >>> if (check_argc_min(argc, 2)) usage(cmd_ready_dev_usage); @@ >>> -293,22 +315,34 @@ static int cmd_ready_dev(int argc, char >>> perror("failed to open /dev/btrfs-control"); return 1; } - if >>> (!is_block_device(argv[1])) { + + path = >>> canonicalize_path(argv[argc - 1]); + if (!path) { >>> fprintf(stderr, - "ERROR: %s is not a block device\n", >>> argv[1]); - close(fd); - return 1; + >>> "ERROR: Could not canonicalize pathname '%s': %s\n", + >>> argv[argc - 1], strerror(errno)); + ret = 1; + goto >>> out; } >>> >>> - strncpy(args.name, argv[argc - 1], BTRFS_PATH_NAME_MAX); + >>> if (!is_block_device(path)) { + fprintf(stderr, + >>> "ERROR: %s is not a block device\n", path); + ret = 1; + >>> goto out; + } + + strncpy(args.name, path, >>> BTRFS_PATH_NAME_MAX); ret = ioctl(fd, BTRFS_IOC_DEVICES_READY, >>> &args); if (ret < 0) { fprintf(stderr, "ERROR: unable to >>> determine if the device '%s'" - " is ready for >>> mounting - %s\n", argv[argc - 1], + " is ready for >>> mounting - %s\n", path, strerror(errno)); ret = 1; } >>> >>> +out: + free(path); close(fd); return ret; } --- >>> a/cmds-replace.c +++ b/cmds-replace.c @@ -134,7 +134,7 @@ static >>> int cmd_start_replace(int argc, c int fddstdev = -1; char *path; >>> char *srcdev; - char *dstdev; + char *dstdev = NULL; int >>> avoid_reading_from_srcdev = 0; int force_using_targetdev = 0; >>> struct stat st; @@ -204,7 +204,12 @@ static int >>> cmd_start_replace(int argc, c } >>> >>> srcdev = argv[optind]; - dstdev = argv[optind + 1]; + >>> dstdev = canonicalize_path(argv[optind + 1]); + if (!dstdev) >>> { + fprintf(stderr, + "ERROR: Could not >>> canonicalize path '%s': %s\n", + argv[optind + 1], >>> strerror(errno)); + } >>> >>> if (is_numerical(srcdev)) { struct btrfs_ioctl_fs_info_args >>> fi_args; @@ -278,6 +283,8 @@ static int cmd_start_replace(int >>> argc, c >>> >>> close(fddstdev); fddstdev = -1; + free(dstdev); + dstdev = >>> NULL; >>> >>> dev_replace_handle_sigint(fdmnt); if (!do_not_background) { @@ >>> -312,6 +319,8 @@ static int cmd_start_replace(int argc, c return >>> 0; >>> >>> leave_with_error: + if (dstdev) + free(dstdev); if >>> (fdmnt != -1) close(fdmnt); if (fdsrcdev != -1) --- a/utils.c +++ >>> b/utils.c @@ -987,6 +987,63 @@ static int >>> blk_file_in_dev_list(struct b } >>> >>> /* + * Resolve a pathname to a device mapper node to >>> /dev/mapper/<name> + * Returns NULL on invalid input or malloc >>> failure; Other failures + * will be handled by the caller using >>> the input pathame. + */ +char *canonicalize_dm_name(const char >>> *ptname) +{ + FILE *f; + size_t sz; + char >>> path[256], name[256], *res = NULL; + + if (!ptname || >>> !*ptname) + return NULL; + + snprintf(path, >>> sizeof(path), "/sys/block/%s/dm/name", ptname); + if (!(f = >>> fopen(path, "r"))) + return NULL; + + /* read <name>\n >>> from sysfs */ + if (fgets(name, sizeof(name), f) && (sz = >>> strlen(name)) > 1) { + name[sz - 1] = '\0'; + >>> snprintf(path, sizeof(path), "/dev/mapper/%s", name); + + >>> if (access(path, F_OK) == 0) + res = strdup(path); + >>> } + fclose(f); + return res; +} + +/* + * Resolve a >>> pathname to a canonical device node, e.g. /dev/sda1 or + * to a >>> device mapper pathname. + * Returns NULL on invalid input or >>> malloc failure; Other failures + * will be handled by the caller >>> using the input pathame. + */ +char *canonicalize_path(const char >>> *path) +{ + char *canonical, *p; + + if (!path || !*path) + >>> return NULL; + + canonical = realpath(path, NULL); + if >>> (!canonical) + return strdup(path); + p = >>> strrchr(canonical, '/'); + if (p && strncmp(p, "/dm-", 4) == 0 >>> && isdigit(*(p + 4))) { + char *dm = >>> canonicalize_dm_name(p + 1); + if (dm) { + >>> free(canonical); + return dm; + } + } + >>> return canonical; +} + +/* * returns 1 if the device was mounted, >>> < 0 on error or 0 if everything * is safe to continue. */ --- >>> a/utils.h +++ b/utils.h @@ -61,6 +61,8 @@ int >>> btrfs_add_to_fsid(struct btrfs_trans int btrfs_scan_for_fsid(int >>> run_ioctls); void btrfs_register_one_device(char *fname); int >>> btrfs_scan_one_dir(char *dirname, int run_ioctl); +char >>> *canonicalize_dm_name(const char *ptname); +char >>> *canonicalize_path(const char *path); int check_mounted(const >>> char *devicename); int check_mounted_where(int fd, const char >>> *file, char *where, int size, struct btrfs_fs_devices >>> **fs_devices_mnt); >>> >> -- To unsubscribe from this list: send the line "unsubscribe >> linux-btrfs" in the body of a message to majordomo@vger.kernel.org >> More majordomo info at http://vger.kernel.org/majordomo-info.html >> > > > - -- > Jeff Mahoney > SUSE Labs > -----BEGIN PGP SIGNATURE----- > Version: GnuPG/MacGPG2 v2.0.22 (Darwin) > > iQIcBAEBAgAGBQJUFhr9AAoJEB57S2MheeWyXzgQAMKR0b29uU0kyYbalOxjpVkB > yAvR5kFItcEgRSfUUyqJqL+atJZ2PrzU4fQsz6i25cOBfQC5EdZYO/QENXe7HleU > oCECYucZC14qTLwswhpMeCtyMrX98sjhQxZ6NPSmLZWi/Btc2QDdwwbij4B3KRnQ > lE/AWmtT/J0fqlmjr18ETdTjTCZVyxUpN809hZpHAwQAiDZhXqbTUveaFu3DpTyn > vDdM5LZ2oGPjS/3WgoNiFy36R1cnxIh3k/+aM/afU20T//dOn8Cdrh+k868hv9fV > f3YfkyoHzwBCXp8VaOZytkOy7HCLbTEpnnB7MbvrNo1ukOlrFZRVRMSUrco2QD6x > 4UYfgBj/zuzxIUZotEnpAmE8ppbMBylqjx+dyIwviDCU0Huw9mEnraBf3Yunuf3k > za92pv/SYJnEFRl2w5ULhtR6GJ3RM8TQJOtWRJjPLkS3L30xvVOQZ1zXoazi/4yX > qp7bKKN/hzC8mPvp8aoCW5c+F652e4pmUplbR4RDaNCD7ipv60XfQBVYOJOGh2YX > +OCYERzFs4th3uPlKNDSfsLHwGY36bPC3mPdiaJW144MX1vHL5YKJiEKcnk5QZNY > BwWidSvdylitAmZeKeCRTOJJh/b+Q7/K72RdWSl6g5qipa/00Gc97Sl/jNOyZ7K6 > URPW75jcC/KtA2gIArtL > =fAyc > -----END PGP SIGNATURE----- > -- > To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html > -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
>>> Nice patch. However its better if we do this in the btrfs kernel >>> function btrfs_scan_one_device(). Since the non-canonicalize path >>> can still sneak through the btrfs specific mount option "device=". >>> >>> Any comments ? >> >> My initial reaction is to avoid playing naming names within the >> kernel. But since it's device mapper-specific, it might not be too >> messy to do that. In addition to the patch to the progs, we're also >> carrying a patch to systemd since it has it's own little ioctl wrapper >> to do the scanning. It needed to be fixed there as well. > > looks like we need to fix systemd as well. As I check systemd is using > READY ioctl. systemd is using device ready ioctl's bug as a feature (if the comment in systemd is true, as below). ./systemd/rules/64-btrfs.rules ---- # let the kernel know about this btrfs filesystem, and check if it is complete IMPORT{builtin}="btrfs ready $devnode" ---- btrfs dev ready is only to check if the device is ready, not to let the kernel know about it. the bug part in the ioctl is it would update the device path, even when the device is mounted. Either we need transition the bug as a feature OR fix the bug. -Anand -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
--- a/cmds-device.c +++ b/cmds-device.c @@ -95,6 +95,7 @@ static int cmd_add_dev(int argc, char ** int devfd, res; u64 dev_block_count = 0; int mixed = 0; + char *path; res = test_dev_for_mkfs(argv[i], force, estr); if (res) { @@ -118,15 +119,24 @@ static int cmd_add_dev(int argc, char ** goto error_out; } - strncpy_null(ioctl_args.name, argv[i]); + path = canonicalize_path(argv[i]); + if (!path) { + fprintf(stderr, + "ERROR: Could not canonicalize pathname '%s': %s\n", + argv[i], strerror(errno)); + ret++; + goto error_out; + } + + strncpy_null(ioctl_args.name, path); res = ioctl(fdmnt, BTRFS_IOC_ADD_DEV, &ioctl_args); e = errno; - if(res<0){ + if (res < 0) { fprintf(stderr, "ERROR: error adding the device '%s' - %s\n", - argv[i], strerror(e)); + path, strerror(e)); ret++; } - + free(path); } error_out: @@ -242,6 +252,7 @@ static int cmd_scan_dev(int argc, char * for( i = devstart ; i < argc ; i++ ){ struct btrfs_ioctl_vol_args args; + char *path; if (!is_block_device(argv[i])) { fprintf(stderr, @@ -249,9 +260,17 @@ static int cmd_scan_dev(int argc, char * ret = 1; goto close_out; } - printf("Scanning for Btrfs filesystems in '%s'\n", argv[i]); + path = canonicalize_path(argv[i]); + if (!path) { + fprintf(stderr, + "ERROR: Could not canonicalize path '%s': %s\n", + argv[i], strerror(errno)); + ret = 1; + goto close_out; + } + printf("Scanning for Btrfs filesystems in '%s'\n", path); - strncpy_null(args.name, argv[i]); + strncpy_null(args.name, path); /* * FIXME: which are the error code returned by this ioctl ? * it seems that is impossible to understand if there no is @@ -262,9 +281,11 @@ static int cmd_scan_dev(int argc, char * if( ret < 0 ){ fprintf(stderr, "ERROR: unable to scan the device '%s' - %s\n", - argv[i], strerror(e)); + path, strerror(e)); + free(path); goto close_out; } + free(path); } close_out: @@ -284,6 +305,7 @@ static int cmd_ready_dev(int argc, char struct btrfs_ioctl_vol_args args; int fd; int ret; + char *path; if (check_argc_min(argc, 2)) usage(cmd_ready_dev_usage); @@ -293,22 +315,34 @@ static int cmd_ready_dev(int argc, char perror("failed to open /dev/btrfs-control"); return 1; } - if (!is_block_device(argv[1])) { + + path = canonicalize_path(argv[argc - 1]); + if (!path) { fprintf(stderr, - "ERROR: %s is not a block device\n", argv[1]); - close(fd); - return 1; + "ERROR: Could not canonicalize pathname '%s': %s\n", + argv[argc - 1], strerror(errno)); + ret = 1; + goto out; } - strncpy(args.name, argv[argc - 1], BTRFS_PATH_NAME_MAX); + if (!is_block_device(path)) { + fprintf(stderr, + "ERROR: %s is not a block device\n", path); + ret = 1; + goto out; + } + + strncpy(args.name, path, BTRFS_PATH_NAME_MAX); ret = ioctl(fd, BTRFS_IOC_DEVICES_READY, &args); if (ret < 0) { fprintf(stderr, "ERROR: unable to determine if the device '%s'" - " is ready for mounting - %s\n", argv[argc - 1], + " is ready for mounting - %s\n", path, strerror(errno)); ret = 1; } +out: + free(path); close(fd); return ret; } --- a/cmds-replace.c +++ b/cmds-replace.c @@ -134,7 +134,7 @@ static int cmd_start_replace(int argc, c int fddstdev = -1; char *path; char *srcdev; - char *dstdev; + char *dstdev = NULL; int avoid_reading_from_srcdev = 0; int force_using_targetdev = 0; struct stat st; @@ -204,7 +204,12 @@ static int cmd_start_replace(int argc, c } srcdev = argv[optind]; - dstdev = argv[optind + 1]; + dstdev = canonicalize_path(argv[optind + 1]); + if (!dstdev) { + fprintf(stderr, + "ERROR: Could not canonicalize path '%s': %s\n", + argv[optind + 1], strerror(errno)); + } if (is_numerical(srcdev)) { struct btrfs_ioctl_fs_info_args fi_args; @@ -278,6 +283,8 @@ static int cmd_start_replace(int argc, c close(fddstdev); fddstdev = -1; + free(dstdev); + dstdev = NULL; dev_replace_handle_sigint(fdmnt); if (!do_not_background) { @@ -312,6 +319,8 @@ static int cmd_start_replace(int argc, c return 0; leave_with_error: + if (dstdev) + free(dstdev); if (fdmnt != -1) close(fdmnt); if (fdsrcdev != -1) --- a/utils.c +++ b/utils.c @@ -987,6 +987,63 @@ static int blk_file_in_dev_list(struct b } /* + * Resolve a pathname to a device mapper node to /dev/mapper/<name> + * Returns NULL on invalid input or malloc failure; Other failures + * will be handled by the caller using the input pathame. + */ +char *canonicalize_dm_name(const char *ptname) +{ + FILE *f; + size_t sz; + char path[256], name[256], *res = NULL; + + if (!ptname || !*ptname) + return NULL; + + snprintf(path, sizeof(path), "/sys/block/%s/dm/name", ptname); + if (!(f = fopen(path, "r"))) + return NULL; + + /* read <name>\n from sysfs */ + if (fgets(name, sizeof(name), f) && (sz = strlen(name)) > 1) { + name[sz - 1] = '\0'; + snprintf(path, sizeof(path), "/dev/mapper/%s", name); + + if (access(path, F_OK) == 0) + res = strdup(path); + } + fclose(f); + return res; +} + +/* + * Resolve a pathname to a canonical device node, e.g. /dev/sda1 or + * to a device mapper pathname. + * Returns NULL on invalid input or malloc failure; Other failures + * will be handled by the caller using the input pathame. + */ +char *canonicalize_path(const char *path) +{ + char *canonical, *p; + + if (!path || !*path) + return NULL; + + canonical = realpath(path, NULL); + if (!canonical) + return strdup(path); + p = strrchr(canonical, '/'); + if (p && strncmp(p, "/dm-", 4) == 0 && isdigit(*(p + 4))) { + char *dm = canonicalize_dm_name(p + 1); + if (dm) { + free(canonical); + return dm; + } + } + return canonical; +} + +/* * returns 1 if the device was mounted, < 0 on error or 0 if everything * is safe to continue. */ --- a/utils.h +++ b/utils.h @@ -61,6 +61,8 @@ int btrfs_add_to_fsid(struct btrfs_trans int btrfs_scan_for_fsid(int run_ioctls); void btrfs_register_one_device(char *fname); int btrfs_scan_one_dir(char *dirname, int run_ioctl); +char *canonicalize_dm_name(const char *ptname); +char *canonicalize_path(const char *path); int check_mounted(const char *devicename); int check_mounted_where(int fd, const char *file, char *where, int size, struct btrfs_fs_devices **fs_devices_mnt);
mount(8) will canonicalize pathnames before passing them to the kernel. Links to e.g. /dev/sda will be resolved to /dev/sda. Links to /dev/dm-# will be resolved using the name of the device mapper table to /dev/mapper/<name>. Btrfs will use whatever name the user passes to it, regardless of whether it is canonical or not. That means that if a 'btrfs device ready' is issued on any device node pointing to the original device, it will adopt the new name instead of the name that was used during mount. Mounting using /dev/sdb2 will result in df: /dev/sdb2 209715200 39328 207577088 1% /mnt # ls -la /dev/whatever-i-like lrwxrwxrwx 1 root root 4 Jun 4 13:36 /dev/whatever-i-like -> sdb2 # btrfs dev ready /dev/whatever-i-like # df /mnt /dev/whatever-i-like 209715200 39328 207577088 1% /mnt Likewise, mounting with /dev/mapper/whatever and using /dev/dm-0 with a btrfs device command results in df showing /dev/dm-0. This can happen with multipath devices with friendly names enabled and doing something like 'partprobe' which (at least with our version) ends up issuing a 'change' uevent on the sysfs node. That *always* uses the dm-# name, and we get confused users. This patch does the same canonicalization of the paths that mount does so that we don't end up having inconsistent names reported by ->show_devices later. Signed-off-by: Jeff Mahoney <jeffm@suse.com> --- cmds-device.c | 60 ++++++++++++++++++++++++++++++++++++++++++++------------- cmds-replace.c | 13 ++++++++++-- utils.c | 57 ++++++++++++++++++++++++++++++++++++++++++++++++++++++ utils.h | 2 + 4 files changed, 117 insertions(+), 15 deletions(-)