| Message ID | 65a82dbc6ea54c5d02a1238f5fa26d19388c5d2a.1652789936.git.jstancek@redhat.com |
|---|---|
| State | Accepted, archived |
| Headers | show |
| Series | [v2] tst_find_backing_dev: match mount point if major/minor can't be found | expand |
Hi! > + len = count_match_len(path, mnt_point); > + if (len > best_match_len) { I wonder if we should also check for minimal len to avoid trival prefix matches such as "/". Maybe we should set the best_match_len to 1 to begin with. > + strcpy(dev, pre); > + best_match_len = len; > + } > } > > SAFE_FCLOSE(NULL, file); > -- > 2.27.0 > > > -- > Mailing list info: https://lists.linux.it/listinfo/ltp
On Tue, May 17, 2022 at 5:30 PM Cyril Hrubis <chrubis@suse.cz> wrote: > > Hi! > > + len = count_match_len(path, mnt_point); > > + if (len > best_match_len) { > > I wonder if we should also check for minimal len to avoid trival prefix > matches such as "/". Maybe we should set the best_match_len to 1 to > begin with. Sure, we can do that. Would you ack this for release as well? This has been failing for quite a while on fedora (which uses btrfs). > > > + strcpy(dev, pre); > > + best_match_len = len; > > + } > > } > > > > SAFE_FCLOSE(NULL, file); > > -- > > 2.27.0 > > > > > > -- > > Mailing list info: https://lists.linux.it/listinfo/ltp > > -- > Cyril Hrubis > chrubis@suse.cz >
Hi! > Sure, we can do that. Would you ack this for release as well? > This has been failing for quite a while on fedora (which uses btrfs). Sure, either way this version looks good. Reviewed-by: Cyril Hrubis <chrubis@suse.cz>
On Thu, May 19, 2022 at 11:59 AM Cyril Hrubis <chrubis@suse.cz> wrote: > > Hi! > > Sure, we can do that. Would you ack this for release as well? > > This has been failing for quite a while on fedora (which uses btrfs). > > Sure, either way this version looks good. > > Reviewed-by: Cyril Hrubis <chrubis@suse.cz> pushed
diff --git a/lib/tst_device.c b/lib/tst_device.c index d296f9118cde..b1af0f0146d4 100644 --- a/lib/tst_device.c +++ b/lib/tst_device.c @@ -506,6 +506,16 @@ unsigned long tst_dev_bytes_written(const char *dev) return dev_bytes_written; } +static int count_match_len(const char *first, const char *second) +{ + int len = 0; + + while (*first && *first++ == *second++) + len++; + + return len; +} + void tst_find_backing_dev(const char *path, char *dev) { struct stat buf; @@ -514,6 +524,8 @@ void tst_find_backing_dev(const char *path, char *dev) char *pre = NULL; char *next = NULL; unsigned int dev_major, dev_minor, line_mjr, line_mnr; + unsigned int len, best_match_len = 0; + char mnt_point[PATH_MAX]; if (stat(path, &buf) < 0) tst_brkm(TWARN | TERRNO, NULL, "stat() failed"); @@ -524,17 +536,25 @@ void tst_find_backing_dev(const char *path, char *dev) *dev = '\0'; while (fgets(line, sizeof(line), file)) { - if (sscanf(line, "%*d %*d %d:%d", &line_mjr, &line_mnr) != 2) + if (sscanf(line, "%*d %*d %d:%d %*s %s", + &line_mjr, &line_mnr, mnt_point) != 3) continue; + pre = strstr(line, " - "); + pre = strtok_r(pre, " ", &next); + pre = strtok_r(NULL, " ", &next); + pre = strtok_r(NULL, " ", &next); + if (line_mjr == dev_major && line_mnr == dev_minor) { - pre = strstr(line, " - "); - pre = strtok_r(pre, " ", &next); - pre = strtok_r(NULL, " ", &next); - pre = strtok_r(NULL, " ", &next); strcpy(dev, pre); break; } + + len = count_match_len(path, mnt_point); + if (len > best_match_len) { + strcpy(dev, pre); + best_match_len = len; + } } SAFE_FCLOSE(NULL, file);
ioctl_loop05 fails on btrfs, because tst_find_backing_dev doesn't find major/minor returned by stat() Per https://lwn.net/Articles/866582 "btrfs allocates a separate device number (the usual major/minor pair) for each subvolume ... and call to on a file within a subvolume will return a device number that does not exist in files like mountinfo." As fallback, if there's no major/minor match, use best match of mount path. Signed-off-by: Jan Stancek <jstancek@redhat.com> --- lib/tst_device.c | 30 +++++++++++++++++++++++++----- 1 file changed, 25 insertions(+), 5 deletions(-)