| Message ID | a8d81afb-8c95-c6bf-028d-8d6e463557b3@redhat.com (mailing list archive) |
|---|---|
| State | Superseded, archived |
| Headers | show |
On Fri, Aug 12, 2016 at 06:07:00PM -0500, Eric Sandeen wrote: > The new XFS_GETQSTATV quotactl, available since kernel v3.12, > was never implemented in xfs_quota, and the "state" command > continues to use XFS_GETQSTAT, which cannot report both > group & project quota on newer formats. > > The new call has room for all 3 quota types (user, group, and > quota), vs just two, where previously project and quota > overlapped. > > So: > > First, try XFS_GETQSTATV. > If it passes, we have all the information we need, and we print > it. state_qfilestat() is modified to take the newer structure. > > If it fails, try XFS_GETQSTAT. If that passes, we are on an > older kernel with neither XFS_GETQSTATV nor the on-disk project > quota inode. We copy the available information into the newer > statv structure, carefully determining wither group or project > (or neither) is actually active, and print it with the same > state_qfilestat routine. > > Signed-off-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Bill O'Donnell <billodo@redhat.com> > --- > > I probably could have done some memcpy()'s in > state_stat_to_statv(), but opted for the explicit copy-out; > the structures aren't identical, although the newer one > only differs by padding on the end. If memcpy() is preferable > I could send a V2... > > > diff --git a/include/xqm.h b/include/xqm.h > index c084b2d..78262c3 100644 > --- a/include/xqm.h > +++ b/include/xqm.h > @@ -32,6 +32,7 @@ > #define Q_XGETQSTAT XQM_CMD(5) /* get quota subsystem status */ > #define Q_XQUOTARM XQM_CMD(6) /* free disk space used by dquots */ > #define Q_XQUOTASYNC XQM_CMD(7) /* delalloc flush, updates dquots */ > +#define Q_XGETQSTATV XQM_CMD(8) /* newer version of get quota */ > #define Q_XGETNEXTQUOTA XQM_CMD(9) /* get disk limits and usage */ > > /* > @@ -149,4 +150,30 @@ typedef struct fs_quota_stat { > __u16 qs_iwarnlimit; /* limit for num warnings */ > } fs_quota_stat_t; > > +/* > + * Some basic information about 'quota files' for Q_XGETQSTATV command > + */ > +struct fs_qfilestatv { > + __u64 qfs_ino; /* inode number */ > + __u64 qfs_nblks; /* number of BBs 512-byte-blks */ > + __u32 qfs_nextents; /* number of extents */ > + __u32 qfs_pad; /* pad for 8-byte alignment */ > +}; > + > +struct fs_quota_statv { > + __s8 qs_version; /* version for future changes */ > + __u8 qs_pad1; /* pad for 16bit alignment */ > + __u16 qs_flags; /* FS_QUOTA_.* flags */ > + __u32 qs_incoredqs; /* number of dquots incore */ > + struct fs_qfilestatv qs_uquota; /* user quota information */ > + struct fs_qfilestatv qs_gquota; /* group quota information */ > + struct fs_qfilestatv qs_pquota; /* project quota information */ > + __s32 qs_btimelimit; /* limit for blks timer */ > + __s32 qs_itimelimit; /* limit for inodes timer */ > + __s32 qs_rtbtimelimit;/* limit for rt blks timer */ > + __u16 qs_bwarnlimit; /* limit for num warnings */ > + __u16 qs_iwarnlimit; /* limit for num warnings */ > + __u64 qs_pad2[8]; /* for future proofing */ > +}; > + > #endif /* __XQM_H__ */ > diff --git a/quota/linux.c b/quota/linux.c > index 74dba01..4f1f3c4 100644 > --- a/quota/linux.c > +++ b/quota/linux.c > @@ -55,6 +55,8 @@ xcommand_to_qcommand( > return Q_XSETQLIM; > case XFS_GETQSTAT: > return Q_XGETQSTAT; > + case XFS_GETQSTATV: > + return Q_XGETQSTATV; > case XFS_QUOTARM: > return Q_XQUOTARM; > case XFS_QSYNC: > diff --git a/quota/state.c b/quota/state.c > index 8186762..5d63579 100644 > --- a/quota/state.c > +++ b/quota/state.c > @@ -111,12 +111,12 @@ remove_help(void) > > static void > state_qfilestat( > - FILE *fp, > - fs_path_t *mount, > - uint type, > - fs_qfilestat_t *qfs, > - int accounting, > - int enforcing) > + FILE *fp, > + struct fs_path *mount, > + uint type, > + struct fs_qfilestatv *qfs, > + int accounting, > + int enforcing) > { > fprintf(fp, _("%s quota state on %s (%s)\n"), type_to_string(type), > mount->fs_dir, mount->fs_name); > @@ -142,39 +142,94 @@ state_timelimit( > time_to_string(timelimit, VERBOSE_FLAG | ABSOLUTE_FLAG)); > } > > +/* > + * fs_quota_stat holds a subset of fs_quota_statv; this copies > + * the smaller into the larger, leaving any not-present fields > + * empty. This is so the same reporting function can be used > + * for both XFS_GETQSTAT and XFS_GETQSTATV results. > + */ > static void > -state_quotafile_mount( > - FILE *fp, > - uint type, > - fs_path_t *mount, > - uint flags) > +state_stat_to_statv( > + struct fs_quota_stat *s, > + struct fs_quota_statv *sv) > { > - fs_quota_stat_t s; > - char *dev = mount->fs_name; > + memset(sv, 0, sizeof(struct fs_quota_statv)); > + > + /* shared information */ > + sv->qs_version = s->qs_version; > + sv->qs_flags = s->qs_flags; > + sv->qs_incoredqs = s->qs_incoredqs; > + sv->qs_btimelimit = s->qs_btimelimit; > + sv->qs_itimelimit = s->qs_itimelimit; > + sv->qs_rtbtimelimit = s->qs_rtbtimelimit; > + sv->qs_bwarnlimit = s->qs_bwarnlimit; > + sv->qs_iwarnlimit = s->qs_iwarnlimit; > + > + /* Always room for uquota */ > + sv->qs_uquota.qfs_ino = s->qs_uquota.qfs_ino; > + sv->qs_uquota.qfs_nblks = s->qs_uquota.qfs_nblks; > + sv->qs_uquota.qfs_nextents = s->qs_uquota.qfs_nextents; > + > + /* > + * If we are here, XFS_GETQSTATV failed and XFS_GETQSTAT passed; > + * that is a very strong hint that we're on a kernel which predates > + * the on-disk pquota inode; both were added in v3.12. So, we do > + * some tricksy determination here. > + * gs_gquota may hold either group quota inode info, or project > + * quota if that is used instead; which one it actually holds depends > + * on the quota flags. (If neither is set, neither is used) > + */ > + if (s->qs_flags & XFS_QUOTA_GDQ_ACCT) { > + /* gs_gquota holds group quota info */ > + sv->qs_gquota.qfs_ino = s->qs_gquota.qfs_ino; > + sv->qs_gquota.qfs_nblks = s->qs_gquota.qfs_nblks; > + sv->qs_gquota.qfs_nextents = s->qs_gquota.qfs_nextents; > + } else if (s->qs_flags & XFS_QUOTA_PDQ_ACCT) { > + /* gs_gquota actually holds project quota info */ > + sv->qs_pquota.qfs_ino = s->qs_gquota.qfs_ino; > + sv->qs_pquota.qfs_nblks = s->qs_gquota.qfs_nblks; > + sv->qs_pquota.qfs_nextents = s->qs_gquota.qfs_nextents; > + } > +} > > - if (xfsquotactl(XFS_GETQSTAT, dev, type, 0, (void *)&s) < 0) { > - if (flags & VERBOSE_FLAG) > - fprintf(fp, _("%s quota are not enabled on %s\n"), > - type_to_string(type), dev); > - return; > +static void > +state_quotafile_mount( > + FILE *fp, > + uint type, > + struct fs_path *mount, > + uint flags) > +{ > + struct fs_quota_stat s; > + struct fs_quota_statv sv; > + char *dev = mount->fs_name; > + > + if (xfsquotactl(XFS_GETQSTATV, dev, type, 0, (void *)&sv) < 0) { > + if (xfsquotactl(XFS_GETQSTAT, dev, type, 0, (void *)&s) < 0) { > + if (flags & VERBOSE_FLAG) > + fprintf(fp, > + _("%s quota are not enabled on %s\n"), > + type_to_string(type), dev); > + return; > + } > + state_stat_to_statv(&s, &sv); > } > > if (type & XFS_USER_QUOTA) > - state_qfilestat(fp, mount, XFS_USER_QUOTA, &s.qs_uquota, > - s.qs_flags & XFS_QUOTA_UDQ_ACCT, > - s.qs_flags & XFS_QUOTA_UDQ_ENFD); > + state_qfilestat(fp, mount, XFS_USER_QUOTA, &sv.qs_uquota, > + sv.qs_flags & XFS_QUOTA_UDQ_ACCT, > + sv.qs_flags & XFS_QUOTA_UDQ_ENFD); > if (type & XFS_GROUP_QUOTA) > - state_qfilestat(fp, mount, XFS_GROUP_QUOTA, &s.qs_gquota, > - s.qs_flags & XFS_QUOTA_GDQ_ACCT, > - s.qs_flags & XFS_QUOTA_GDQ_ENFD); > + state_qfilestat(fp, mount, XFS_GROUP_QUOTA, &sv.qs_gquota, > + sv.qs_flags & XFS_QUOTA_GDQ_ACCT, > + sv.qs_flags & XFS_QUOTA_GDQ_ENFD); > if (type & XFS_PROJ_QUOTA) > - state_qfilestat(fp, mount, XFS_PROJ_QUOTA, &s.qs_gquota, > - s.qs_flags & XFS_QUOTA_PDQ_ACCT, > - s.qs_flags & XFS_QUOTA_PDQ_ENFD); > + state_qfilestat(fp, mount, XFS_PROJ_QUOTA, &sv.qs_pquota, > + sv.qs_flags & XFS_QUOTA_PDQ_ACCT, > + sv.qs_flags & XFS_QUOTA_PDQ_ENFD); > > - state_timelimit(fp, XFS_BLOCK_QUOTA, s.qs_btimelimit); > - state_timelimit(fp, XFS_INODE_QUOTA, s.qs_itimelimit); > - state_timelimit(fp, XFS_RTBLOCK_QUOTA, s.qs_rtbtimelimit); > + state_timelimit(fp, XFS_BLOCK_QUOTA, sv.qs_btimelimit); > + state_timelimit(fp, XFS_INODE_QUOTA, sv.qs_itimelimit); > + state_timelimit(fp, XFS_RTBLOCK_QUOTA, sv.qs_rtbtimelimit); > } > > static void > > > _______________________________________________ > xfs mailing list > xfs@oss.sgi.com > http://oss.sgi.com/mailman/listinfo/xfs
On Fri, Aug 12, 2016 at 06:07:00PM -0500, Eric Sandeen wrote: > The new XFS_GETQSTATV quotactl, available since kernel v3.12, > was never implemented in xfs_quota, and the "state" command > continues to use XFS_GETQSTAT, which cannot report both > group & project quota on newer formats. > > The new call has room for all 3 quota types (user, group, and > quota), vs just two, where previously project and quota > overlapped. > > So: > > First, try XFS_GETQSTATV. > If it passes, we have all the information we need, and we print > it. state_qfilestat() is modified to take the newer structure. > > If it fails, try XFS_GETQSTAT. If that passes, we are on an > older kernel with neither XFS_GETQSTATV nor the on-disk project > quota inode. We copy the available information into the newer > statv structure, carefully determining wither group or project > (or neither) is actually active, and print it with the same > state_qfilestat routine. > > Signed-off-by: Eric Sandeen <sandeen@redhat.com> > --- > > I probably could have done some memcpy()'s in > state_stat_to_statv(), but opted for the explicit copy-out; > the structures aren't identical, although the newer one > only differs by padding on the end. If memcpy() is preferable > I could send a V2... > > > diff --git a/include/xqm.h b/include/xqm.h > index c084b2d..78262c3 100644 > --- a/include/xqm.h > +++ b/include/xqm.h > @@ -32,6 +32,7 @@ > #define Q_XGETQSTAT XQM_CMD(5) /* get quota subsystem status */ > #define Q_XQUOTARM XQM_CMD(6) /* free disk space used by dquots */ > #define Q_XQUOTASYNC XQM_CMD(7) /* delalloc flush, updates dquots */ > +#define Q_XGETQSTATV XQM_CMD(8) /* newer version of get quota */ > #define Q_XGETNEXTQUOTA XQM_CMD(9) /* get disk limits and usage */ > > /* > @@ -149,4 +150,30 @@ typedef struct fs_quota_stat { > __u16 qs_iwarnlimit; /* limit for num warnings */ > } fs_quota_stat_t; > > +/* > + * Some basic information about 'quota files' for Q_XGETQSTATV command > + */ > +struct fs_qfilestatv { > + __u64 qfs_ino; /* inode number */ > + __u64 qfs_nblks; /* number of BBs 512-byte-blks */ > + __u32 qfs_nextents; /* number of extents */ > + __u32 qfs_pad; /* pad for 8-byte alignment */ > +}; > + > +struct fs_quota_statv { > + __s8 qs_version; /* version for future changes */ > + __u8 qs_pad1; /* pad for 16bit alignment */ > + __u16 qs_flags; /* FS_QUOTA_.* flags */ > + __u32 qs_incoredqs; /* number of dquots incore */ > + struct fs_qfilestatv qs_uquota; /* user quota information */ > + struct fs_qfilestatv qs_gquota; /* group quota information */ > + struct fs_qfilestatv qs_pquota; /* project quota information */ > + __s32 qs_btimelimit; /* limit for blks timer */ > + __s32 qs_itimelimit; /* limit for inodes timer */ > + __s32 qs_rtbtimelimit;/* limit for rt blks timer */ > + __u16 qs_bwarnlimit; /* limit for num warnings */ > + __u16 qs_iwarnlimit; /* limit for num warnings */ > + __u64 qs_pad2[8]; /* for future proofing */ > +}; > + > #endif /* __XQM_H__ */ > diff --git a/quota/linux.c b/quota/linux.c > index 74dba01..4f1f3c4 100644 > --- a/quota/linux.c > +++ b/quota/linux.c > @@ -55,6 +55,8 @@ xcommand_to_qcommand( > return Q_XSETQLIM; > case XFS_GETQSTAT: > return Q_XGETQSTAT; > + case XFS_GETQSTATV: > + return Q_XGETQSTATV; > case XFS_QUOTARM: > return Q_XQUOTARM; > case XFS_QSYNC: > diff --git a/quota/state.c b/quota/state.c > index 8186762..5d63579 100644 > --- a/quota/state.c > +++ b/quota/state.c > @@ -111,12 +111,12 @@ remove_help(void) > > static void > state_qfilestat( > - FILE *fp, > - fs_path_t *mount, > - uint type, > - fs_qfilestat_t *qfs, > - int accounting, > - int enforcing) > + FILE *fp, > + struct fs_path *mount, > + uint type, > + struct fs_qfilestatv *qfs, > + int accounting, > + int enforcing) > { > fprintf(fp, _("%s quota state on %s (%s)\n"), type_to_string(type), > mount->fs_dir, mount->fs_name); > @@ -142,39 +142,94 @@ state_timelimit( > time_to_string(timelimit, VERBOSE_FLAG | ABSOLUTE_FLAG)); > } > > +/* > + * fs_quota_stat holds a subset of fs_quota_statv; this copies > + * the smaller into the larger, leaving any not-present fields > + * empty. This is so the same reporting function can be used > + * for both XFS_GETQSTAT and XFS_GETQSTATV results. > + */ > static void > -state_quotafile_mount( > - FILE *fp, > - uint type, > - fs_path_t *mount, > - uint flags) > +state_stat_to_statv( > + struct fs_quota_stat *s, > + struct fs_quota_statv *sv) > { > - fs_quota_stat_t s; > - char *dev = mount->fs_name; > + memset(sv, 0, sizeof(struct fs_quota_statv)); > + > + /* shared information */ > + sv->qs_version = s->qs_version; > + sv->qs_flags = s->qs_flags; > + sv->qs_incoredqs = s->qs_incoredqs; > + sv->qs_btimelimit = s->qs_btimelimit; > + sv->qs_itimelimit = s->qs_itimelimit; > + sv->qs_rtbtimelimit = s->qs_rtbtimelimit; > + sv->qs_bwarnlimit = s->qs_bwarnlimit; > + sv->qs_iwarnlimit = s->qs_iwarnlimit; > + > + /* Always room for uquota */ > + sv->qs_uquota.qfs_ino = s->qs_uquota.qfs_ino; > + sv->qs_uquota.qfs_nblks = s->qs_uquota.qfs_nblks; > + sv->qs_uquota.qfs_nextents = s->qs_uquota.qfs_nextents; > + > + /* > + * If we are here, XFS_GETQSTATV failed and XFS_GETQSTAT passed; > + * that is a very strong hint that we're on a kernel which predates > + * the on-disk pquota inode; both were added in v3.12. So, we do > + * some tricksy determination here. > + * gs_gquota may hold either group quota inode info, or project > + * quota if that is used instead; which one it actually holds depends > + * on the quota flags. (If neither is set, neither is used) > + */ > + if (s->qs_flags & XFS_QUOTA_GDQ_ACCT) { > + /* gs_gquota holds group quota info */ > + sv->qs_gquota.qfs_ino = s->qs_gquota.qfs_ino; > + sv->qs_gquota.qfs_nblks = s->qs_gquota.qfs_nblks; > + sv->qs_gquota.qfs_nextents = s->qs_gquota.qfs_nextents; > + } else if (s->qs_flags & XFS_QUOTA_PDQ_ACCT) { > + /* gs_gquota actually holds project quota info */ > + sv->qs_pquota.qfs_ino = s->qs_gquota.qfs_ino; > + sv->qs_pquota.qfs_nblks = s->qs_gquota.qfs_nblks; > + sv->qs_pquota.qfs_nextents = s->qs_gquota.qfs_nextents; > + } > +} > > - if (xfsquotactl(XFS_GETQSTAT, dev, type, 0, (void *)&s) < 0) { > - if (flags & VERBOSE_FLAG) > - fprintf(fp, _("%s quota are not enabled on %s\n"), > - type_to_string(type), dev); > - return; > +static void > +state_quotafile_mount( > + FILE *fp, > + uint type, > + struct fs_path *mount, > + uint flags) > +{ > + struct fs_quota_stat s; > + struct fs_quota_statv sv; > + char *dev = mount->fs_name; > + When I tested on XFS with crc=1, I got below result: [root@dhcp-13-149 xfsprogs-dev]# mount /dev/mapper/testvg-scratchdev /mnt/scratch -o gquota,pquota [root@dhcp-13-149 xfsprogs-dev]# xfs_quota -xc "state" /mnt/scratch User quota state on /mnt/scratch (/dev/mapper/testvg-scratchdev) Accounting: OFF Enforcement: OFF Inode: #101 (1 blocks, 1 extents) Group quota state on /mnt/scratch (/dev/mapper/testvg-scratchdev) Accounting: ON Enforcement: ON Inode: #99 (1 blocks, 1 extents) Project quota state on /mnt/scratch (/dev/mapper/testvg-scratchdev) Accounting: ON Enforcement: ON Inode: #0 (0 blocks, 0 extents) Blocks grace time: [7 days] Inodes grace time: [7 days] Realtime Blocks grace time: [7 days] I got project quota inode number is #0, when I mount gquota and pquota together. That's incorrect, because: [root@dhcp-13-149 xfsprogs-dev]# xfs_db -r -c "sb 0" -c "p" /dev/mapper/testvg-scratchdev|grep quot uquotino = 101 gquotino = 99 pquotino = 100 From the result of strace, we can see: quotactl(0x5808 /* Q_??? */|USRQUOTA, "/dev/mapper/testvg-scratchdev", 0, 0x7ffdbbda94c0) = -1 EINVAL (Invalid argument) quotactl(Q_XGETQSTAT|USRQUOTA, "/dev/mapper/testvg-scratchdev", 0, {version=1, ...}) = 0 So the first call for XFS_GETQSTATV return EINVAL. In linux kernel quota_getxstatev() function, the logic is: /* If this kernel doesn't support user specified version, fail */ switch (fqs.qs_version) { case FS_QSTATV_VERSION1: break; default: return -EINVAL; } So we need to set qa_version to FS_QSTATV_VERSION1, before we call XFS_GETQSTATV. And as we talked on IRC, it test passed if I set sv.qs_version = 1; at here. So maybe a V2 patch is needed to fix this problem:) Thanks, Zorro > + if (xfsquotactl(XFS_GETQSTATV, dev, type, 0, (void *)&sv) < 0) { > + if (xfsquotactl(XFS_GETQSTAT, dev, type, 0, (void *)&s) < 0) { > + if (flags & VERBOSE_FLAG) > + fprintf(fp, > + _("%s quota are not enabled on %s\n"), > + type_to_string(type), dev); > + return; > + } > + state_stat_to_statv(&s, &sv); > } > > if (type & XFS_USER_QUOTA) > - state_qfilestat(fp, mount, XFS_USER_QUOTA, &s.qs_uquota, > - s.qs_flags & XFS_QUOTA_UDQ_ACCT, > - s.qs_flags & XFS_QUOTA_UDQ_ENFD); > + state_qfilestat(fp, mount, XFS_USER_QUOTA, &sv.qs_uquota, > + sv.qs_flags & XFS_QUOTA_UDQ_ACCT, > + sv.qs_flags & XFS_QUOTA_UDQ_ENFD); > if (type & XFS_GROUP_QUOTA) > - state_qfilestat(fp, mount, XFS_GROUP_QUOTA, &s.qs_gquota, > - s.qs_flags & XFS_QUOTA_GDQ_ACCT, > - s.qs_flags & XFS_QUOTA_GDQ_ENFD); > + state_qfilestat(fp, mount, XFS_GROUP_QUOTA, &sv.qs_gquota, > + sv.qs_flags & XFS_QUOTA_GDQ_ACCT, > + sv.qs_flags & XFS_QUOTA_GDQ_ENFD); > if (type & XFS_PROJ_QUOTA) > - state_qfilestat(fp, mount, XFS_PROJ_QUOTA, &s.qs_gquota, > - s.qs_flags & XFS_QUOTA_PDQ_ACCT, > - s.qs_flags & XFS_QUOTA_PDQ_ENFD); > + state_qfilestat(fp, mount, XFS_PROJ_QUOTA, &sv.qs_pquota, > + sv.qs_flags & XFS_QUOTA_PDQ_ACCT, > + sv.qs_flags & XFS_QUOTA_PDQ_ENFD); > > - state_timelimit(fp, XFS_BLOCK_QUOTA, s.qs_btimelimit); > - state_timelimit(fp, XFS_INODE_QUOTA, s.qs_itimelimit); > - state_timelimit(fp, XFS_RTBLOCK_QUOTA, s.qs_rtbtimelimit); > + state_timelimit(fp, XFS_BLOCK_QUOTA, sv.qs_btimelimit); > + state_timelimit(fp, XFS_INODE_QUOTA, sv.qs_itimelimit); > + state_timelimit(fp, XFS_RTBLOCK_QUOTA, sv.qs_rtbtimelimit); > } > > static void > > > _______________________________________________ > xfs mailing list > xfs@oss.sgi.com > http://oss.sgi.com/mailman/listinfo/xfs
On 8/15/16 11:38 AM, Zorro Lang wrote: > From the result of strace, we can see: > quotactl(0x5808 /* Q_??? */|USRQUOTA, "/dev/mapper/testvg-scratchdev", 0, 0x7ffdbbda94c0) = -1 EINVAL (Invalid argument) > quotactl(Q_XGETQSTAT|USRQUOTA, "/dev/mapper/testvg-scratchdev", 0, {version=1, ...}) = 0 > > So the first call for XFS_GETQSTATV return EINVAL. > > In linux kernel quota_getxstatev() function, the logic is: > /* If this kernel doesn't support user specified version, fail */ > switch (fqs.qs_version) { > case FS_QSTATV_VERSION1: > break; > default: > return -EINVAL; > } > > So we need to set qa_version to FS_QSTATV_VERSION1, before > we call XFS_GETQSTATV. And as we talked on IRC, it test passed if I set > > sv.qs_version = 1; > > at here. > > So maybe a V2 patch is needed to fix this problem:) Yep, kernel commit said: | Callers of the new quotactl command have to set the version of the data | structure being passed, and kernel will fill as much data as requested. | If the kernel does not support the user-space provided version, EINVAL | will be returned. User-space can reduce the version number and call the same | quotactl again. but this isn't documented anywhere else, AFAICT; this call still isn't documented in the quotactl manpage. o_O I'll fix it up, thanks for catching it! -Eric
diff --git a/include/xqm.h b/include/xqm.h index c084b2d..78262c3 100644 --- a/include/xqm.h +++ b/include/xqm.h @@ -32,6 +32,7 @@ #define Q_XGETQSTAT XQM_CMD(5) /* get quota subsystem status */ #define Q_XQUOTARM XQM_CMD(6) /* free disk space used by dquots */ #define Q_XQUOTASYNC XQM_CMD(7) /* delalloc flush, updates dquots */ +#define Q_XGETQSTATV XQM_CMD(8) /* newer version of get quota */ #define Q_XGETNEXTQUOTA XQM_CMD(9) /* get disk limits and usage */ /* @@ -149,4 +150,30 @@ typedef struct fs_quota_stat { __u16 qs_iwarnlimit; /* limit for num warnings */ } fs_quota_stat_t; +/* + * Some basic information about 'quota files' for Q_XGETQSTATV command + */ +struct fs_qfilestatv { + __u64 qfs_ino; /* inode number */ + __u64 qfs_nblks; /* number of BBs 512-byte-blks */ + __u32 qfs_nextents; /* number of extents */ + __u32 qfs_pad; /* pad for 8-byte alignment */ +}; + +struct fs_quota_statv { + __s8 qs_version; /* version for future changes */ + __u8 qs_pad1; /* pad for 16bit alignment */ + __u16 qs_flags; /* FS_QUOTA_.* flags */ + __u32 qs_incoredqs; /* number of dquots incore */ + struct fs_qfilestatv qs_uquota; /* user quota information */ + struct fs_qfilestatv qs_gquota; /* group quota information */ + struct fs_qfilestatv qs_pquota; /* project quota information */ + __s32 qs_btimelimit; /* limit for blks timer */ + __s32 qs_itimelimit; /* limit for inodes timer */ + __s32 qs_rtbtimelimit;/* limit for rt blks timer */ + __u16 qs_bwarnlimit; /* limit for num warnings */ + __u16 qs_iwarnlimit; /* limit for num warnings */ + __u64 qs_pad2[8]; /* for future proofing */ +}; + #endif /* __XQM_H__ */ diff --git a/quota/linux.c b/quota/linux.c index 74dba01..4f1f3c4 100644 --- a/quota/linux.c +++ b/quota/linux.c @@ -55,6 +55,8 @@ xcommand_to_qcommand( return Q_XSETQLIM; case XFS_GETQSTAT: return Q_XGETQSTAT; + case XFS_GETQSTATV: + return Q_XGETQSTATV; case XFS_QUOTARM: return Q_XQUOTARM; case XFS_QSYNC: diff --git a/quota/state.c b/quota/state.c index 8186762..5d63579 100644 --- a/quota/state.c +++ b/quota/state.c @@ -111,12 +111,12 @@ remove_help(void) static void state_qfilestat( - FILE *fp, - fs_path_t *mount, - uint type, - fs_qfilestat_t *qfs, - int accounting, - int enforcing) + FILE *fp, + struct fs_path *mount, + uint type, + struct fs_qfilestatv *qfs, + int accounting, + int enforcing) { fprintf(fp, _("%s quota state on %s (%s)\n"), type_to_string(type), mount->fs_dir, mount->fs_name); @@ -142,39 +142,94 @@ state_timelimit( time_to_string(timelimit, VERBOSE_FLAG | ABSOLUTE_FLAG)); } +/* + * fs_quota_stat holds a subset of fs_quota_statv; this copies + * the smaller into the larger, leaving any not-present fields + * empty. This is so the same reporting function can be used + * for both XFS_GETQSTAT and XFS_GETQSTATV results. + */ static void -state_quotafile_mount( - FILE *fp, - uint type, - fs_path_t *mount, - uint flags) +state_stat_to_statv( + struct fs_quota_stat *s, + struct fs_quota_statv *sv) { - fs_quota_stat_t s; - char *dev = mount->fs_name; + memset(sv, 0, sizeof(struct fs_quota_statv)); + + /* shared information */ + sv->qs_version = s->qs_version; + sv->qs_flags = s->qs_flags; + sv->qs_incoredqs = s->qs_incoredqs; + sv->qs_btimelimit = s->qs_btimelimit; + sv->qs_itimelimit = s->qs_itimelimit; + sv->qs_rtbtimelimit = s->qs_rtbtimelimit; + sv->qs_bwarnlimit = s->qs_bwarnlimit; + sv->qs_iwarnlimit = s->qs_iwarnlimit; + + /* Always room for uquota */ + sv->qs_uquota.qfs_ino = s->qs_uquota.qfs_ino; + sv->qs_uquota.qfs_nblks = s->qs_uquota.qfs_nblks; + sv->qs_uquota.qfs_nextents = s->qs_uquota.qfs_nextents; + + /* + * If we are here, XFS_GETQSTATV failed and XFS_GETQSTAT passed; + * that is a very strong hint that we're on a kernel which predates + * the on-disk pquota inode; both were added in v3.12. So, we do + * some tricksy determination here. + * gs_gquota may hold either group quota inode info, or project + * quota if that is used instead; which one it actually holds depends + * on the quota flags. (If neither is set, neither is used) + */ + if (s->qs_flags & XFS_QUOTA_GDQ_ACCT) { + /* gs_gquota holds group quota info */ + sv->qs_gquota.qfs_ino = s->qs_gquota.qfs_ino; + sv->qs_gquota.qfs_nblks = s->qs_gquota.qfs_nblks; + sv->qs_gquota.qfs_nextents = s->qs_gquota.qfs_nextents; + } else if (s->qs_flags & XFS_QUOTA_PDQ_ACCT) { + /* gs_gquota actually holds project quota info */ + sv->qs_pquota.qfs_ino = s->qs_gquota.qfs_ino; + sv->qs_pquota.qfs_nblks = s->qs_gquota.qfs_nblks; + sv->qs_pquota.qfs_nextents = s->qs_gquota.qfs_nextents; + } +} - if (xfsquotactl(XFS_GETQSTAT, dev, type, 0, (void *)&s) < 0) { - if (flags & VERBOSE_FLAG) - fprintf(fp, _("%s quota are not enabled on %s\n"), - type_to_string(type), dev); - return; +static void +state_quotafile_mount( + FILE *fp, + uint type, + struct fs_path *mount, + uint flags) +{ + struct fs_quota_stat s; + struct fs_quota_statv sv; + char *dev = mount->fs_name; + + if (xfsquotactl(XFS_GETQSTATV, dev, type, 0, (void *)&sv) < 0) { + if (xfsquotactl(XFS_GETQSTAT, dev, type, 0, (void *)&s) < 0) { + if (flags & VERBOSE_FLAG) + fprintf(fp, + _("%s quota are not enabled on %s\n"), + type_to_string(type), dev); + return; + } + state_stat_to_statv(&s, &sv); } if (type & XFS_USER_QUOTA) - state_qfilestat(fp, mount, XFS_USER_QUOTA, &s.qs_uquota, - s.qs_flags & XFS_QUOTA_UDQ_ACCT, - s.qs_flags & XFS_QUOTA_UDQ_ENFD); + state_qfilestat(fp, mount, XFS_USER_QUOTA, &sv.qs_uquota, + sv.qs_flags & XFS_QUOTA_UDQ_ACCT, + sv.qs_flags & XFS_QUOTA_UDQ_ENFD); if (type & XFS_GROUP_QUOTA) - state_qfilestat(fp, mount, XFS_GROUP_QUOTA, &s.qs_gquota, - s.qs_flags & XFS_QUOTA_GDQ_ACCT, - s.qs_flags & XFS_QUOTA_GDQ_ENFD); + state_qfilestat(fp, mount, XFS_GROUP_QUOTA, &sv.qs_gquota, + sv.qs_flags & XFS_QUOTA_GDQ_ACCT, + sv.qs_flags & XFS_QUOTA_GDQ_ENFD); if (type & XFS_PROJ_QUOTA) - state_qfilestat(fp, mount, XFS_PROJ_QUOTA, &s.qs_gquota, - s.qs_flags & XFS_QUOTA_PDQ_ACCT, - s.qs_flags & XFS_QUOTA_PDQ_ENFD); + state_qfilestat(fp, mount, XFS_PROJ_QUOTA, &sv.qs_pquota, + sv.qs_flags & XFS_QUOTA_PDQ_ACCT, + sv.qs_flags & XFS_QUOTA_PDQ_ENFD); - state_timelimit(fp, XFS_BLOCK_QUOTA, s.qs_btimelimit); - state_timelimit(fp, XFS_INODE_QUOTA, s.qs_itimelimit); - state_timelimit(fp, XFS_RTBLOCK_QUOTA, s.qs_rtbtimelimit); + state_timelimit(fp, XFS_BLOCK_QUOTA, sv.qs_btimelimit); + state_timelimit(fp, XFS_INODE_QUOTA, sv.qs_itimelimit); + state_timelimit(fp, XFS_RTBLOCK_QUOTA, sv.qs_rtbtimelimit); } static void
The new XFS_GETQSTATV quotactl, available since kernel v3.12, was never implemented in xfs_quota, and the "state" command continues to use XFS_GETQSTAT, which cannot report both group & project quota on newer formats. The new call has room for all 3 quota types (user, group, and quota), vs just two, where previously project and quota overlapped. So: First, try XFS_GETQSTATV. If it passes, we have all the information we need, and we print it. state_qfilestat() is modified to take the newer structure. If it fails, try XFS_GETQSTAT. If that passes, we are on an older kernel with neither XFS_GETQSTATV nor the on-disk project quota inode. We copy the available information into the newer statv structure, carefully determining wither group or project (or neither) is actually active, and print it with the same state_qfilestat routine. Signed-off-by: Eric Sandeen <sandeen@redhat.com> --- I probably could have done some memcpy()'s in state_stat_to_statv(), but opted for the explicit copy-out; the structures aren't identical, although the newer one only differs by padding on the end. If memcpy() is preferable I could send a V2...