* [PATCH 0/4] xfsprogs: bulkstat v5 @ 2019-09-06 3:35 Darrick J. Wong 2019-09-06 3:35 ` [PATCH 1/4] xfs_io: add a bulkstat command Darrick J. Wong ` (3 more replies) 0 siblings, 4 replies; 11+ messages in thread From: Darrick J. Wong @ 2019-09-06 3:35 UTC (permalink / raw) To: sandeen, darrick.wong; +Cc: linux-xfs Hi all, In this series, we do the heavier lifting to support the new v5 bulkstat code. First, we add an xfs_io command so that we can actually test the old and new interfaces. Then, we start converting spaceman and scrub to use the per-AG bulkstat functions so that they can shed a lot of duplicated geometry calculation code. Finally, we speed up scrub phase 7 by fixing a minor thinko. If you're going to start using this mess, you probably ought to just pull from my git trees, which are linked below. This is an extraordinary way to destroy everything. Enjoy! Comments and questions are, as always, welcome. --D xfsprogs git tree: https://git.kernel.org/cgit/linux/kernel/git/djwong/xfsprogs-dev.git/log/?h=bulkstat-v5-factoring ^ permalink raw reply [flat|nested] 11+ messages in thread
* [PATCH 1/4] xfs_io: add a bulkstat command 2019-09-06 3:35 [PATCH 0/4] xfsprogs: bulkstat v5 Darrick J. Wong @ 2019-09-06 3:35 ` Darrick J. Wong 2019-09-12 23:51 ` Dave Chinner 2019-09-06 3:35 ` [PATCH 2/4] xfs_spaceman: remove open-coded per-ag bulkstat Darrick J. Wong ` (2 subsequent siblings) 3 siblings, 1 reply; 11+ messages in thread From: Darrick J. Wong @ 2019-09-06 3:35 UTC (permalink / raw) To: sandeen, darrick.wong; +Cc: linux-xfs From: Darrick J. Wong <darrick.wong@oracle.com> Add a bulkstat command to xfs_io so that we can test our new xfrog code. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> --- io/Makefile | 9 - io/bulkstat.c | 533 ++++++++++++++++++++++++++++++++++++++++++++++++++++ io/init.c | 1 io/io.h | 1 libfrog/bulkstat.c | 20 ++ libfrog/bulkstat.h | 3 man/man8/xfs_io.8 | 68 +++++++ 7 files changed, 631 insertions(+), 4 deletions(-) create mode 100644 io/bulkstat.c diff --git a/io/Makefile b/io/Makefile index 484e2b5a..1112605e 100644 --- a/io/Makefile +++ b/io/Makefile @@ -9,10 +9,11 @@ LTCOMMAND = xfs_io LSRCFILES = xfs_bmap.sh xfs_freeze.sh xfs_mkfile.sh HFILES = init.h io.h CFILES = init.c \ - attr.c bmap.c crc32cselftest.c cowextsize.c encrypt.c file.c freeze.c \ - fsync.c getrusage.c imap.c inject.c label.c link.c mmap.c open.c \ - parent.c pread.c prealloc.c pwrite.c reflink.c resblks.c scrub.c \ - seek.c shutdown.c stat.c swapext.c sync.c truncate.c utimes.c + attr.c bmap.c bulkstat.c crc32cselftest.c cowextsize.c encrypt.c \ + file.c freeze.c fsync.c getrusage.c imap.c inject.c label.c link.c \ + mmap.c open.c parent.c pread.c prealloc.c pwrite.c reflink.c \ + resblks.c scrub.c seek.c shutdown.c stat.c swapext.c sync.c \ + truncate.c utimes.c LLDLIBS = $(LIBXCMD) $(LIBHANDLE) $(LIBFROG) $(LIBPTHREAD) LTDEPENDENCIES = $(LIBXCMD) $(LIBHANDLE) $(LIBFROG) diff --git a/io/bulkstat.c b/io/bulkstat.c new file mode 100644 index 00000000..76ba682b --- /dev/null +++ b/io/bulkstat.c @@ -0,0 +1,533 @@ +// SPDX-License-Identifier: GPL-2.0+ +/* + * Copyright (C) 2019 Oracle. All Rights Reserved. + * Author: Darrick J. Wong <darrick.wong@oracle.com> + */ +#include "xfs.h" +#include "platform_defs.h" +#include "command.h" +#include "init.h" +#include "libfrog/fsgeom.h" +#include "libfrog/bulkstat.h" +#include "libfrog/paths.h" +#include "io.h" +#include "input.h" + +static bool debug; + +static void +dump_bulkstat_time( + const char *tag, + uint64_t sec, + uint32_t nsec) +{ + printf("\t%s = %"PRIu64".%"PRIu32"\n", tag, sec, nsec); +} + +static void +dump_bulkstat( + struct xfs_bulkstat *bstat) +{ + printf("bs_ino = %"PRIu64"\n", bstat->bs_ino); + printf("\tbs_size = %"PRIu64"\n", bstat->bs_size); + + printf("\tbs_blocks = %"PRIu64"\n", bstat->bs_blocks); + printf("\tbs_xflags = 0x%"PRIx64"\n", bstat->bs_xflags); + + dump_bulkstat_time("bs_atime", bstat->bs_atime, bstat->bs_atime_nsec); + dump_bulkstat_time("bs_ctime", bstat->bs_ctime, bstat->bs_ctime_nsec); + dump_bulkstat_time("bs_mtime", bstat->bs_mtime, bstat->bs_mtime_nsec); + dump_bulkstat_time("bs_btime", bstat->bs_btime, bstat->bs_btime_nsec); + + printf("\tbs_gen = 0x%"PRIx32"\n", bstat->bs_gen); + printf("\tbs_uid = %"PRIu32"\n", bstat->bs_uid); + printf("\tbs_gid = %"PRIu32"\n", bstat->bs_gid); + printf("\tbs_projectid = %"PRIu32"\n", bstat->bs_projectid); + + printf("\tbs_blksize = %"PRIu32"\n", bstat->bs_blksize); + printf("\tbs_rdev = %"PRIu32"\n", bstat->bs_rdev); + printf("\tbs_cowextsize_blks = %"PRIu32"\n", bstat->bs_cowextsize_blks); + printf("\tbs_extsize_blks = %"PRIu32"\n", bstat->bs_extsize_blks); + + printf("\tbs_nlink = %"PRIu32"\n", bstat->bs_nlink); + printf("\tbs_extents = %"PRIu32"\n", bstat->bs_extents); + printf("\tbs_aextents = %"PRIu32"\n", bstat->bs_aextents); + printf("\tbs_version = %"PRIu16"\n", bstat->bs_version); + printf("\tbs_forkoff = %"PRIu16"\n", bstat->bs_forkoff); + + printf("\tbs_sick = 0x%"PRIx16"\n", bstat->bs_sick); + printf("\tbs_checked = 0x%"PRIx16"\n", bstat->bs_checked); + printf("\tbs_mode = 0%"PRIo16"\n", bstat->bs_mode); +}; + +static void +bulkstat_help(void) +{ + printf(_( +"Bulk-queries the filesystem for inode stat information and prints it.\n" +"\n" +" -a Only iterate this AG.\n" +" -d Print debugging output.\n" +" -e Stop after this inode.\n" +" -n Ask for this many results at once.\n" +" -s Inode to start with.\n" +" -v Use this version of the ioctl (1 or 5).\n")); +} + +static int +bulkstat_f( + int argc, + char **argv) +{ + struct xfs_fd xfd = XFS_FD_INIT(file->fd); + struct xfs_bulkstat_req *breq; + unsigned long long startino = 0; + unsigned long long endino = -1ULL; + unsigned long batch_size = 4096; + unsigned long agno = 0; + unsigned long ver = 0; + bool has_agno = false; + unsigned int i; + int c; + int ret; + + while ((c = getopt(argc, argv, "a:cde:n:qs:v:")) != -1) { + switch (c) { + case 'a': + errno = 0; + agno = strtoul(optarg, NULL, 10); + if (errno) { + perror(optarg); + return 1; + } + has_agno = true; + break; + case 'd': + debug = true; + break; + case 'e': + errno = 0; + endino = strtoull(optarg, NULL, 10); + if (errno) { + perror(optarg); + return 1; + } + break; + case 'n': + errno = 0; + batch_size = strtoul(optarg, NULL, 10); + if (errno) { + perror(optarg); + return 1; + } + break; + case 's': + errno = 0; + startino = strtoull(optarg, NULL, 10); + if (errno) { + perror(optarg); + return 1; + } + break; + case 'v': + errno = 0; + ver = strtoull(optarg, NULL, 10); + if (errno) { + perror(optarg); + return 1; + } + if (ver != 1 && ver != 5) { + fprintf(stderr, "version must be 1 or 5.\n"); + return 1; + } + break; + default: + bulkstat_help(); + return 0; + } + } + if (optind != argc) { + bulkstat_help(); + return 0; + } + + ret = xfd_prepare_geometry(&xfd); + if (ret) { + errno = ret; + perror("xfd_prepare_geometry"); + exitcode = 1; + return 0; + } + + breq = xfrog_bulkstat_alloc_req(batch_size, startino); + if (!breq) { + perror("alloc bulkreq"); + exitcode = 1; + return 0; + } + + if (has_agno) + xfrog_bulkstat_set_ag(breq, agno); + + switch (ver) { + case 1: + xfd.flags |= XFROG_FLAG_BULKSTAT_FORCE_V1; + break; + case 5: + xfd.flags |= XFROG_FLAG_BULKSTAT_FORCE_V5; + break; + default: + break; + } + + while ((ret = xfrog_bulkstat(&xfd, breq)) == 0) { + if (debug) + printf( +_("bulkstat: startino=%lld flags=0x%x agno=%u ret=%d icount=%u ocount=%u\n"), + (long long)breq->hdr.ino, + (unsigned int)breq->hdr.flags, + (unsigned int)breq->hdr.agno, + ret, + (unsigned int)breq->hdr.icount, + (unsigned int)breq->hdr.ocount); + if (breq->hdr.ocount == 0) + break; + + for (i = 0; i < breq->hdr.ocount; i++) { + if (breq->bulkstat[i].bs_ino > endino) + break; + dump_bulkstat(&breq->bulkstat[i]); + } + } + if (ret) { + errno = ret; + perror("xfrog_bulkstat"); + exitcode = 1; + return 0; + } + + free(breq); + return 0; +} + +static void +bulkstat_single_help(void) +{ + printf(_( +"Queries the filesystem for a single inode's stat information and prints it.\n" +"\n" +" -v Use this version of the ioctl (1 or 5).\n" +"\n" +"Pass in inode numbers or a special inode name:\n" +" root Root directory.\n")); +} + +struct single_map { + const char *tag; + uint64_t code; +}; + +struct single_map tags[] = { + {"root", XFS_BULK_IREQ_SPECIAL_ROOT}, + {NULL, 0}, +}; + +static int +bulkstat_single_f( + int argc, + char **argv) +{ + struct xfs_fd xfd = XFS_FD_INIT(file->fd); + struct xfs_bulkstat bulkstat; + unsigned long ver = 0; + unsigned int i; + int c; + int ret; + + while ((c = getopt(argc, argv, "v:")) != -1) { + switch (c) { + case 'v': + errno = 0; + ver = strtoull(optarg, NULL, 10); + if (errno) { + perror(optarg); + return 1; + } + if (ver != 1 && ver != 5) { + fprintf(stderr, "version must be 1 or 5.\n"); + return 1; + } + break; + default: + bulkstat_single_help(); + return 0; + } + } + + ret = xfd_prepare_geometry(&xfd); + if (ret) { + errno = ret; + perror("xfd_prepare_geometry"); + exitcode = 1; + return 0; + } + + switch (ver) { + case 1: + xfd.flags |= XFROG_FLAG_BULKSTAT_FORCE_V1; + break; + case 5: + xfd.flags |= XFROG_FLAG_BULKSTAT_FORCE_V5; + break; + default: + break; + } + + for (i = optind; i < argc; i++) { + struct single_map *sm = tags; + uint64_t ino; + unsigned int flags = 0; + + /* Try to look up our tag... */ + for (sm = tags; sm->tag; sm++) { + if (!strcmp(argv[i], sm->tag)) { + ino = sm->code; + flags |= XFS_BULK_IREQ_SPECIAL; + break; + } + } + + /* ...or else it's an inode number. */ + if (sm->tag == NULL) { + errno = 0; + ino = strtoull(argv[i], NULL, 10); + if (errno) { + perror(argv[i]); + exitcode = 1; + return 0; + } + } + + ret = xfrog_bulkstat_single(&xfd, ino, flags, &bulkstat); + if (ret) { + errno = ret; + perror("xfrog_bulkstat_single"); + continue; + } + + if (debug) + printf( +_("bulkstat_single: startino=%"PRIu64" flags=0x%"PRIx32" ret=%d\n"), + ino, flags, ret); + + dump_bulkstat(&bulkstat); + } + + return 0; +} + +static void +dump_inumbers( + struct xfs_inumbers *inumbers) +{ + printf("xi_startino = %"PRIu64"\n", inumbers->xi_startino); + printf("\txi_allocmask = 0x%"PRIx64"\n", inumbers->xi_allocmask); + printf("\txi_alloccount = %"PRIu8"\n", inumbers->xi_alloccount); + printf("\txi_version = %"PRIu8"\n", inumbers->xi_version); +} + +static void +inumbers_help(void) +{ + printf(_( +"Queries the filesystem for inode group information and prints it.\n" +"\n" +" -a Only iterate this AG.\n" +" -d Print debugging output.\n" +" -e Stop after this inode.\n" +" -n Ask for this many results at once.\n" +" -s Inode to start with.\n" +" -v Use this version of the ioctl (1 or 5).\n")); +} + +static int +inumbers_f( + int argc, + char **argv) +{ + struct xfs_fd xfd = XFS_FD_INIT(file->fd); + struct xfs_inumbers_req *ireq; + unsigned long long startino = 0; + unsigned long long endino = -1ULL; + unsigned long batch_size = 4096; + unsigned long agno = 0; + unsigned long ver = 0; + bool has_agno = false; + unsigned int i; + int c; + int ret; + + while ((c = getopt(argc, argv, "a:cde:n:qs:v:")) != -1) { + switch (c) { + case 'a': + errno = 0; + agno = strtoul(optarg, NULL, 10); + if (errno) { + perror(optarg); + return 1; + } + has_agno = true; + break; + case 'd': + debug = true; + break; + case 'e': + errno = 0; + endino = strtoull(optarg, NULL, 10); + if (errno) { + perror(optarg); + return 1; + } + break; + case 'n': + errno = 0; + batch_size = strtoul(optarg, NULL, 10); + if (errno) { + perror(optarg); + return 1; + } + break; + case 's': + errno = 0; + startino = strtoull(optarg, NULL, 10); + if (errno) { + perror(optarg); + return 1; + } + break; + case 'v': + errno = 0; + ver = strtoull(optarg, NULL, 10); + if (errno) { + perror(optarg); + return 1; + } + if (ver != 1 && ver != 5) { + fprintf(stderr, "version must be 1 or 5.\n"); + return 1; + } + break; + default: + bulkstat_help(); + return 0; + } + } + if (optind != argc) { + bulkstat_help(); + return 0; + } + + ret = xfd_prepare_geometry(&xfd); + if (ret) { + errno = ret; + perror("xfd_prepare_geometry"); + exitcode = 1; + return 0; + } + + ireq = xfrog_inumbers_alloc_req(batch_size, startino); + if (!ireq) { + perror("alloc inumbersreq"); + exitcode = 1; + return 0; + } + + if (has_agno) + xfrog_inumbers_set_ag(ireq, agno); + + switch (ver) { + case 1: + xfd.flags |= XFROG_FLAG_BULKSTAT_FORCE_V1; + break; + case 5: + xfd.flags |= XFROG_FLAG_BULKSTAT_FORCE_V5; + break; + default: + break; + } + + while ((ret = xfrog_inumbers(&xfd, ireq)) == 0) { + if (debug) + printf( +_("bulkstat: startino=%"PRIu64" flags=0x%"PRIx32" agno=%"PRIu32" ret=%d icount=%"PRIu32" ocount=%"PRIu32"\n"), + ireq->hdr.ino, + ireq->hdr.flags, + ireq->hdr.agno, + ret, + ireq->hdr.icount, + ireq->hdr.ocount); + if (ireq->hdr.ocount == 0) + break; + + for (i = 0; i < ireq->hdr.ocount; i++) { + if (ireq->inumbers[i].xi_startino > endino) + break; + dump_inumbers(&ireq->inumbers[i]); + } + } + if (ret) { + errno = ret; + perror("xfrog_inumbers"); + exitcode = 1; + return 0; + } + + free(ireq); + return 0; +} + +static cmdinfo_t bulkstat_cmd = { + .name = "bulkstat", + .cfunc = bulkstat_f, + .argmin = 0, + .argmax = -1, + .flags = CMD_NOMAP_OK, + .help = bulkstat_help, +}; + +static cmdinfo_t bulkstat_single_cmd = { + .name = "bulkstat_single", + .cfunc = bulkstat_single_f, + .argmin = 0, + .argmax = -1, + .flags = CMD_NOMAP_OK, + .help = bulkstat_single_help, +}; + +static cmdinfo_t inumbers_cmd = { + .name = "inumbers", + .cfunc = inumbers_f, + .argmin = 0, + .argmax = -1, + .flags = CMD_NOMAP_OK, + .help = inumbers_help, +}; + +void +bulkstat_init(void) +{ + bulkstat_cmd.args = + _("[-a agno] [-d] [-e endino] [-n batchsize] [-s startino]"); + bulkstat_cmd.oneline = _("Bulk stat of inodes in a filesystem"); + + bulkstat_single_cmd.args = _("inum..."); + bulkstat_single_cmd.oneline = _("Stat one inode in a filesystem"); + + inumbers_cmd.args = + _("[-a agno] [-d] [-e endino] [-n batchsize] [-s startino]"); + inumbers_cmd.oneline = _("Query inode groups in a filesystem"); + + add_command(&bulkstat_cmd); + add_command(&bulkstat_single_cmd); + add_command(&inumbers_cmd); +} diff --git a/io/init.c b/io/init.c index 7025aea5..033ed67d 100644 --- a/io/init.c +++ b/io/init.c @@ -46,6 +46,7 @@ init_commands(void) { attr_init(); bmap_init(); + bulkstat_init(); copy_range_init(); cowextsize_init(); encrypt_init(); diff --git a/io/io.h b/io/io.h index 00dff2b7..49db902f 100644 --- a/io/io.h +++ b/io/io.h @@ -183,3 +183,4 @@ extern void log_writes_init(void); extern void scrub_init(void); extern void repair_init(void); extern void crc32cselftest_init(void); +extern void bulkstat_init(void); diff --git a/libfrog/bulkstat.c b/libfrog/bulkstat.c index 748d0f32..603a9589 100644 --- a/libfrog/bulkstat.c +++ b/libfrog/bulkstat.c @@ -387,6 +387,16 @@ xfrog_bulkstat_alloc_req( return breq; } +/* Set a bulkstat cursor to iterate only a particular AG. */ +void +xfrog_bulkstat_set_ag( + struct xfs_bulkstat_req *req, + uint32_t agno) +{ + req->hdr.agno = agno; + req->hdr.flags |= XFS_BULK_IREQ_AGNO; +} + /* Convert an inumbers (v5) struct to a inogrp (v1) struct. */ void xfrog_inumbers_to_inogrp( @@ -514,3 +524,13 @@ xfrog_inumbers_alloc_req( return ireq; } + +/* Set an inumbers cursor to iterate only a particular AG. */ +void +xfrog_inumbers_set_ag( + struct xfs_inumbers_req *req, + uint32_t agno) +{ + req->hdr.agno = agno; + req->hdr.flags |= XFS_BULK_IREQ_AGNO; +} diff --git a/libfrog/bulkstat.h b/libfrog/bulkstat.h index 5da7d3f5..bed4ff15 100644 --- a/libfrog/bulkstat.h +++ b/libfrog/bulkstat.h @@ -19,11 +19,14 @@ void xfrog_bulkstat_to_bstat(struct xfs_fd *xfd, struct xfs_bstat *bs1, void xfrog_bstat_to_bulkstat(struct xfs_fd *xfd, struct xfs_bulkstat *bstat, const struct xfs_bstat *bs1); +void xfrog_bulkstat_set_ag(struct xfs_bulkstat_req *req, uint32_t agno); + struct xfs_inogrp; int xfrog_inumbers(struct xfs_fd *xfd, struct xfs_inumbers_req *req); struct xfs_inumbers_req *xfrog_inumbers_alloc_req(uint32_t nr, uint64_t startino); +void xfrog_inumbers_set_ag(struct xfs_inumbers_req *req, uint32_t agno); void xfrog_inumbers_to_inogrp(struct xfs_inogrp *ig1, const struct xfs_inumbers *ig); void xfrog_inogrp_to_inumbers(struct xfs_inumbers *ig, diff --git a/man/man8/xfs_io.8 b/man/man8/xfs_io.8 index 6e064bdd..8fd3ffbe 100644 --- a/man/man8/xfs_io.8 +++ b/man/man8/xfs_io.8 @@ -996,6 +996,45 @@ for the current memory mapping. .SH FILESYSTEM COMMANDS .TP +.BI "bulkstat [ \-a " agno " ] [ \-d ] [ \-e " endino " ] [ \-n " batchsize " ] [ \-s " startino " ] +Display raw stat information about a bunch of inodes in an XFS filesystem. +Options are as follows: +.RS 1.0i +.PD 0 +.TP +.BI \-a " agno" +Display only results from the given allocation group. +Defaults to zero. +.TP +.BI \-d +Print debugging information about call results. +.TP +.BI \-e " endino" +Stop displaying records when this inode number is reached. +Defaults to stopping when the system call stops returning results. +.TP +.BI \-n " batchsize" +Retrieve at most this many records per call. +.TP +.BI \-s " startino" +Display inode allocation records starting with this inode. +Defaults to zero. +If this value is zero, then display starts with the allocation group +given above. +.RE +.PD +.TP +.BI "bulkstat_single [ " inum... " | " special... " ] +Display raw stat information about individual inodes in an XFS filesystem. +Arguments must be inode numbers or any of the special values: +.RS 1.0i +.PD 0 +.TP +.B root +Display information about the root directory inode. +.RE +.PD +.TP .B freeze Suspend all write I/O requests to the filesystem of the current file. Only available in expert mode and requires privileges. @@ -1067,6 +1106,35 @@ was specified on the command line, the maximum possible inode number in the system will be printed along with its size. .PD .TP +.BI "inumbers [ \-a " agno " ] [ \-d ] [ \-e " endino " ] [ \-n " batchsize " ] [ \-s " startino " ] +Prints allocation information about groups of inodes in an XFS filesystem. +Callers can use this information to figure out which inodes are allocated. +Options are as follows: +.RS 1.0i +.PD 0 +.TP +.BI \-a " agno" +Display only results from the given allocation group. +Defaults to zero. +.TP +.BI \-d +Print debugging information about call results. +.TP +.BI \-e " endino" +Stop displaying records when this inode number is reached. +Defaults to stopping when the system call stops returning results. +.TP +.BI \-n " batchsize" +Retrieve at most this many records per call. +.TP +.BI \-s " startino" +Display inode allocation records starting with this inode. +Defaults to zero. +If this value is zero, then display starts with the allocation group +given above. +.RE +.PD +.TP .BI "scrub " type " [ " agnumber " | " "ino" " " "gen" " ]" Scrub internal XFS filesystem metadata. The .BI type ^ permalink raw reply related [flat|nested] 11+ messages in thread
* Re: [PATCH 1/4] xfs_io: add a bulkstat command 2019-09-06 3:35 ` [PATCH 1/4] xfs_io: add a bulkstat command Darrick J. Wong @ 2019-09-12 23:51 ` Dave Chinner 0 siblings, 0 replies; 11+ messages in thread From: Dave Chinner @ 2019-09-12 23:51 UTC (permalink / raw) To: Darrick J. Wong; +Cc: sandeen, linux-xfs On Thu, Sep 05, 2019 at 08:35:44PM -0700, Darrick J. Wong wrote: > From: Darrick J. Wong <darrick.wong@oracle.com> > > Add a bulkstat command to xfs_io so that we can test our new xfrog code. > > Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> > --- > io/Makefile | 9 - > io/bulkstat.c | 533 ++++++++++++++++++++++++++++++++++++++++++++++++++++ > io/init.c | 1 > io/io.h | 1 > libfrog/bulkstat.c | 20 ++ > libfrog/bulkstat.h | 3 > man/man8/xfs_io.8 | 68 +++++++ > 7 files changed, 631 insertions(+), 4 deletions(-) > create mode 100644 io/bulkstat.c > > > diff --git a/io/Makefile b/io/Makefile > index 484e2b5a..1112605e 100644 > --- a/io/Makefile > +++ b/io/Makefile > @@ -9,10 +9,11 @@ LTCOMMAND = xfs_io > LSRCFILES = xfs_bmap.sh xfs_freeze.sh xfs_mkfile.sh > HFILES = init.h io.h > CFILES = init.c \ > - attr.c bmap.c crc32cselftest.c cowextsize.c encrypt.c file.c freeze.c \ > - fsync.c getrusage.c imap.c inject.c label.c link.c mmap.c open.c \ > - parent.c pread.c prealloc.c pwrite.c reflink.c resblks.c scrub.c \ > - seek.c shutdown.c stat.c swapext.c sync.c truncate.c utimes.c > + attr.c bmap.c bulkstat.c crc32cselftest.c cowextsize.c encrypt.c \ > + file.c freeze.c fsync.c getrusage.c imap.c inject.c label.c link.c \ > + mmap.c open.c parent.c pread.c prealloc.c pwrite.c reflink.c \ > + resblks.c scrub.c seek.c shutdown.c stat.c swapext.c sync.c \ > + truncate.c utimes.c > > LLDLIBS = $(LIBXCMD) $(LIBHANDLE) $(LIBFROG) $(LIBPTHREAD) > LTDEPENDENCIES = $(LIBXCMD) $(LIBHANDLE) $(LIBFROG) > diff --git a/io/bulkstat.c b/io/bulkstat.c > new file mode 100644 > index 00000000..76ba682b > --- /dev/null > +++ b/io/bulkstat.c > @@ -0,0 +1,533 @@ > +// SPDX-License-Identifier: GPL-2.0+ > +/* > + * Copyright (C) 2019 Oracle. All Rights Reserved. > + * Author: Darrick J. Wong <darrick.wong@oracle.com> > + */ > +#include "xfs.h" > +#include "platform_defs.h" > +#include "command.h" > +#include "init.h" > +#include "libfrog/fsgeom.h" > +#include "libfrog/bulkstat.h" > +#include "libfrog/paths.h" > +#include "io.h" > +#include "input.h" > + > +static bool debug; > + > +static void > +dump_bulkstat_time( > + const char *tag, > + uint64_t sec, > + uint32_t nsec) > +{ > + printf("\t%s = %"PRIu64".%"PRIu32"\n", tag, sec, nsec); > +} > + > +static void > +dump_bulkstat( > + struct xfs_bulkstat *bstat) > +{ > + printf("bs_ino = %"PRIu64"\n", bstat->bs_ino); > + printf("\tbs_size = %"PRIu64"\n", bstat->bs_size); > + > + printf("\tbs_blocks = %"PRIu64"\n", bstat->bs_blocks); > + printf("\tbs_xflags = 0x%"PRIx64"\n", bstat->bs_xflags); > + > + dump_bulkstat_time("bs_atime", bstat->bs_atime, bstat->bs_atime_nsec); > + dump_bulkstat_time("bs_ctime", bstat->bs_ctime, bstat->bs_ctime_nsec); > + dump_bulkstat_time("bs_mtime", bstat->bs_mtime, bstat->bs_mtime_nsec); > + dump_bulkstat_time("bs_btime", bstat->bs_btime, bstat->bs_btime_nsec); > + > + printf("\tbs_gen = 0x%"PRIx32"\n", bstat->bs_gen); > + printf("\tbs_uid = %"PRIu32"\n", bstat->bs_uid); > + printf("\tbs_gid = %"PRIu32"\n", bstat->bs_gid); > + printf("\tbs_projectid = %"PRIu32"\n", bstat->bs_projectid); > + > + printf("\tbs_blksize = %"PRIu32"\n", bstat->bs_blksize); > + printf("\tbs_rdev = %"PRIu32"\n", bstat->bs_rdev); > + printf("\tbs_cowextsize_blks = %"PRIu32"\n", bstat->bs_cowextsize_blks); > + printf("\tbs_extsize_blks = %"PRIu32"\n", bstat->bs_extsize_blks); > + > + printf("\tbs_nlink = %"PRIu32"\n", bstat->bs_nlink); > + printf("\tbs_extents = %"PRIu32"\n", bstat->bs_extents); > + printf("\tbs_aextents = %"PRIu32"\n", bstat->bs_aextents); > + printf("\tbs_version = %"PRIu16"\n", bstat->bs_version); > + printf("\tbs_forkoff = %"PRIu16"\n", bstat->bs_forkoff); > + > + printf("\tbs_sick = 0x%"PRIx16"\n", bstat->bs_sick); > + printf("\tbs_checked = 0x%"PRIx16"\n", bstat->bs_checked); > + printf("\tbs_mode = 0%"PRIo16"\n", bstat->bs_mode); > +}; > + > +static void > +bulkstat_help(void) > +{ > + printf(_( > +"Bulk-queries the filesystem for inode stat information and prints it.\n" > +"\n" > +" -a Only iterate this AG.\n" > +" -d Print debugging output.\n" > +" -e Stop after this inode.\n" > +" -n Ask for this many results at once.\n" > +" -s Inode to start with.\n" > +" -v Use this version of the ioctl (1 or 5).\n")); > +} > + > +static int > +bulkstat_f( > + int argc, > + char **argv) > +{ > + struct xfs_fd xfd = XFS_FD_INIT(file->fd); > + struct xfs_bulkstat_req *breq; > + unsigned long long startino = 0; > + unsigned long long endino = -1ULL; > + unsigned long batch_size = 4096; > + unsigned long agno = 0; > + unsigned long ver = 0; > + bool has_agno = false; > + unsigned int i; > + int c; > + int ret; > + > + while ((c = getopt(argc, argv, "a:cde:n:qs:v:")) != -1) { options c and q are not documented above, and not handled in the case statement below. > + switch (c) { > + case 'a': > + errno = 0; > + agno = strtoul(optarg, NULL, 10); > + if (errno) { > + perror(optarg); > + return 1; > + } > + has_agno = true; > + break; Why not use cvt_u32() and friends for these so they are directly converted to the correct type and overflow checked at the same time? [...] > +static void > +inumbers_help(void) > +{ > + printf(_( > +"Queries the filesystem for inode group information and prints it.\n" > +"\n" > +" -a Only iterate this AG.\n" > +" -d Print debugging output.\n" > +" -e Stop after this inode.\n" > +" -n Ask for this many results at once.\n" > +" -s Inode to start with.\n" > +" -v Use this version of the ioctl (1 or 5).\n")); > +} > + > +static int > +inumbers_f( > + int argc, > + char **argv) > +{ > + struct xfs_fd xfd = XFS_FD_INIT(file->fd); > + struct xfs_inumbers_req *ireq; > + unsigned long long startino = 0; > + unsigned long long endino = -1ULL; > + unsigned long batch_size = 4096; > + unsigned long agno = 0; > + unsigned long ver = 0; > + bool has_agno = false; > + unsigned int i; > + int c; > + int ret; > + > + while ((c = getopt(argc, argv, "a:cde:n:qs:v:")) != -1) { Again, c and q not defined. Same comments about cvt_type() as well... > + if (has_agno) > + xfrog_inumbers_set_ag(ireq, agno); > + > + switch (ver) { > + case 1: > + xfd.flags |= XFROG_FLAG_BULKSTAT_FORCE_V1; > + break; > + case 5: > + xfd.flags |= XFROG_FLAG_BULKSTAT_FORCE_V5; > + break; > + default: > + break; > + } Common helper? > +static cmdinfo_t bulkstat_cmd = { > + .name = "bulkstat", > + .cfunc = bulkstat_f, > + .argmin = 0, > + .argmax = -1, > + .flags = CMD_NOMAP_OK, > + .help = bulkstat_help, > +}; Theres are all oneshot commands, right? > + > +static cmdinfo_t bulkstat_single_cmd = { > + .name = "bulkstat_single", > + .cfunc = bulkstat_single_f, > + .argmin = 0, > + .argmax = -1, > + .flags = CMD_NOMAP_OK, > + .help = bulkstat_single_help, > +}; Doesn't this require at least one parameter? > > .SH FILESYSTEM COMMANDS > .TP > +.BI "bulkstat [ \-a " agno " ] [ \-d ] [ \-e " endino " ] [ \-n " batchsize " ] [ \-s " startino " ] > +Display raw stat information about a bunch of inodes in an XFS filesystem. > +Options are as follows: > +.RS 1.0i > +.PD 0 > +.TP > +.BI \-a " agno" > +Display only results from the given allocation group. > +Defaults to zero. Need to say "If not specified, will all AGs in the fielsystem" > @@ -1067,6 +1106,35 @@ was specified on the command line, the maximum possible inode number in > the system will be printed along with its size. > .PD > .TP > +.BI "inumbers [ \-a " agno " ] [ \-d ] [ \-e " endino " ] [ \-n " batchsize " ] [ \-s " startino " ] > +Prints allocation information about groups of inodes in an XFS filesystem. > +Callers can use this information to figure out which inodes are allocated. > +Options are as follows: > +.RS 1.0i > +.PD 0 > +.TP > +.BI \-a " agno" > +Display only results from the given allocation group. > +Defaults to zero. Same again. Cheers, Dave. -- Dave Chinner david@fromorbit.com ^ permalink raw reply [flat|nested] 11+ messages in thread
* [PATCH 2/4] xfs_spaceman: remove open-coded per-ag bulkstat 2019-09-06 3:35 [PATCH 0/4] xfsprogs: bulkstat v5 Darrick J. Wong 2019-09-06 3:35 ` [PATCH 1/4] xfs_io: add a bulkstat command Darrick J. Wong @ 2019-09-06 3:35 ` Darrick J. Wong 2019-09-12 23:52 ` Dave Chinner 2019-09-06 3:36 ` [PATCH 3/4] xfs_scrub: convert to per-ag inode bulkstat operations Darrick J. Wong 2019-09-06 3:36 ` [PATCH 4/4] xfs_scrub: batch inumbers calls during fscounters calculation Darrick J. Wong 3 siblings, 1 reply; 11+ messages in thread From: Darrick J. Wong @ 2019-09-06 3:35 UTC (permalink / raw) To: sandeen, darrick.wong; +Cc: linux-xfs From: Darrick J. Wong <darrick.wong@oracle.com> Now that xfrog_bulkstat supports per-AG bulkstat, we can get rid of the open coded one in spaceman. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> --- spaceman/health.c | 16 +++++----------- 1 file changed, 5 insertions(+), 11 deletions(-) diff --git a/spaceman/health.c b/spaceman/health.c index b6e1fcd9..e70fd6fd 100644 --- a/spaceman/health.c +++ b/spaceman/health.c @@ -268,28 +268,22 @@ report_bulkstat_health( { struct xfs_bulkstat_req *breq; char descr[256]; - uint64_t startino = 0; - uint64_t lastino = -1ULL; uint32_t i; int error; - if (agno != NULLAGNUMBER) { - startino = cvt_agino_to_ino(&file->xfd, agno, 0); - lastino = cvt_agino_to_ino(&file->xfd, agno + 1, 0) - 1; - } - - breq = xfrog_bulkstat_alloc_req(BULKSTAT_NR, startino); + breq = xfrog_bulkstat_alloc_req(BULKSTAT_NR, 0); if (!breq) { perror("bulk alloc req"); exitcode = 1; return 1; } + if (agno != NULLAGNUMBER) + xfrog_bulkstat_set_ag(breq, agno); + while ((error = xfrog_bulkstat(&file->xfd, breq) == 0) && breq->hdr.ocount > 0) { for (i = 0; i < breq->hdr.ocount; i++) { - if (breq->bulkstat[i].bs_ino > lastino) - goto out; snprintf(descr, sizeof(descr) - 1, _("inode %"PRIu64), breq->bulkstat[i].bs_ino); report_sick(descr, inode_flags, @@ -301,7 +295,7 @@ report_bulkstat_health( errno = error; perror("bulkstat"); } -out: + free(breq); return error; } ^ permalink raw reply related [flat|nested] 11+ messages in thread
* Re: [PATCH 2/4] xfs_spaceman: remove open-coded per-ag bulkstat 2019-09-06 3:35 ` [PATCH 2/4] xfs_spaceman: remove open-coded per-ag bulkstat Darrick J. Wong @ 2019-09-12 23:52 ` Dave Chinner 0 siblings, 0 replies; 11+ messages in thread From: Dave Chinner @ 2019-09-12 23:52 UTC (permalink / raw) To: Darrick J. Wong; +Cc: sandeen, linux-xfs On Thu, Sep 05, 2019 at 08:35:54PM -0700, Darrick J. Wong wrote: > From: Darrick J. Wong <darrick.wong@oracle.com> > > Now that xfrog_bulkstat supports per-AG bulkstat, we can get rid of the > open coded one in spaceman. > > Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Looks good. Reviewed-by: Dave Chinner <dchinner@redhat.com> -- Dave Chinner david@fromorbit.com ^ permalink raw reply [flat|nested] 11+ messages in thread
* [PATCH 3/4] xfs_scrub: convert to per-ag inode bulkstat operations 2019-09-06 3:35 [PATCH 0/4] xfsprogs: bulkstat v5 Darrick J. Wong 2019-09-06 3:35 ` [PATCH 1/4] xfs_io: add a bulkstat command Darrick J. Wong 2019-09-06 3:35 ` [PATCH 2/4] xfs_spaceman: remove open-coded per-ag bulkstat Darrick J. Wong @ 2019-09-06 3:36 ` Darrick J. Wong 2019-09-12 23:55 ` Dave Chinner 2019-09-06 3:36 ` [PATCH 4/4] xfs_scrub: batch inumbers calls during fscounters calculation Darrick J. Wong 3 siblings, 1 reply; 11+ messages in thread From: Darrick J. Wong @ 2019-09-06 3:36 UTC (permalink / raw) To: sandeen, darrick.wong; +Cc: linux-xfs From: Darrick J. Wong <darrick.wong@oracle.com> Now that we're done reworking the xfrog bulkstat wrapper functions, we can adapt xfs_scrub to use the per-ag iteration functionality. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> --- scrub/fscounters.c | 22 ++++++---------------- scrub/inodes.c | 20 ++++++-------------- 2 files changed, 12 insertions(+), 30 deletions(-) diff --git a/scrub/fscounters.c b/scrub/fscounters.c index 2fdf658a..a2cf8171 100644 --- a/scrub/fscounters.c +++ b/scrub/fscounters.c @@ -35,29 +35,25 @@ struct xfs_count_inodes { * exist in the filesystem, assuming we've already scrubbed that. */ static bool -xfs_count_inodes_range( +xfs_count_inodes_ag( struct scrub_ctx *ctx, const char *descr, - uint64_t first_ino, - uint64_t last_ino, + uint32_t agno, uint64_t *count) { struct xfs_inumbers_req *ireq; uint64_t nr = 0; int error; - ASSERT(!(first_ino & (XFS_INODES_PER_CHUNK - 1))); - ASSERT((last_ino & (XFS_INODES_PER_CHUNK - 1))); - - ireq = xfrog_inumbers_alloc_req(1, first_ino); + ireq = xfrog_inumbers_alloc_req(1, 0); if (!ireq) { str_info(ctx, descr, _("Insufficient memory; giving up.")); return false; } + xfrog_inumbers_set_ag(ireq, agno); while (!(error = xfrog_inumbers(&ctx->mnt, ireq))) { - if (ireq->hdr.ocount == 0 || - ireq->inumbers[0].xi_startino >= last_ino) + if (ireq->hdr.ocount == 0) break; nr += ireq->inumbers[0].xi_alloccount; } @@ -83,8 +79,6 @@ xfs_count_ag_inodes( struct xfs_count_inodes *ci = arg; struct scrub_ctx *ctx = (struct scrub_ctx *)wq->wq_ctx; char descr[DESCR_BUFSZ]; - uint64_t ag_ino; - uint64_t next_ag_ino; bool moveon; snprintf(descr, DESCR_BUFSZ, _("dev %d:%d AG %u inodes"), @@ -92,11 +86,7 @@ xfs_count_ag_inodes( minor(ctx->fsinfo.fs_datadev), agno); - ag_ino = cvt_agino_to_ino(&ctx->mnt, agno, 0); - next_ag_ino = cvt_agino_to_ino(&ctx->mnt, agno + 1, 0); - - moveon = xfs_count_inodes_range(ctx, descr, ag_ino, next_ag_ino - 1, - &ci->counters[agno]); + moveon = xfs_count_inodes_ag(ctx, descr, agno, &ci->counters[agno]); if (!moveon) ci->moveon = false; } diff --git a/scrub/inodes.c b/scrub/inodes.c index 65c404ab..c7aadae7 100644 --- a/scrub/inodes.c +++ b/scrub/inodes.c @@ -82,12 +82,11 @@ xfs_iterate_inodes_range_check( * but we also can detect iget failures. */ static bool -xfs_iterate_inodes_range( +xfs_iterate_inodes_ag( struct scrub_ctx *ctx, const char *descr, void *fshandle, - uint64_t first_ino, - uint64_t last_ino, + uint32_t agno, xfs_inode_iter_fn fn, void *arg) { @@ -113,13 +112,14 @@ xfs_iterate_inodes_range( return false; } - ireq = xfrog_inumbers_alloc_req(1, first_ino); + ireq = xfrog_inumbers_alloc_req(1, 0); if (!ireq) { str_info(ctx, descr, _("Insufficient memory; giving up.")); free(breq); return false; } inogrp = &ireq->inumbers[0]; + xfrog_inumbers_set_ag(ireq, agno); /* Find the inode chunk & alloc mask */ error = xfrog_inumbers(&ctx->mnt, ireq); @@ -147,9 +147,6 @@ xfs_iterate_inodes_range( for (i = 0, bs = breq->bulkstat; i < inogrp->xi_alloccount; i++, bs++) { - if (bs->bs_ino > last_ino) - goto out; - handle.ha_fid.fid_ino = bs->bs_ino; handle.ha_fid.fid_gen = bs->bs_gen; error = fn(ctx, &handle, bs, arg); @@ -214,8 +211,6 @@ xfs_scan_ag_inodes( struct xfs_scan_inodes *si = arg; struct scrub_ctx *ctx = (struct scrub_ctx *)wq->wq_ctx; char descr[DESCR_BUFSZ]; - uint64_t ag_ino; - uint64_t next_ag_ino; bool moveon; snprintf(descr, DESCR_BUFSZ, _("dev %d:%d AG %u inodes"), @@ -223,11 +218,8 @@ xfs_scan_ag_inodes( minor(ctx->fsinfo.fs_datadev), agno); - ag_ino = cvt_agino_to_ino(&ctx->mnt, agno, 0); - next_ag_ino = cvt_agino_to_ino(&ctx->mnt, agno + 1, 0); - - moveon = xfs_iterate_inodes_range(ctx, descr, ctx->fshandle, ag_ino, - next_ag_ino - 1, si->fn, si->arg); + moveon = xfs_iterate_inodes_ag(ctx, descr, ctx->fshandle, agno, + si->fn, si->arg); if (!moveon) si->moveon = false; } ^ permalink raw reply related [flat|nested] 11+ messages in thread
* Re: [PATCH 3/4] xfs_scrub: convert to per-ag inode bulkstat operations 2019-09-06 3:36 ` [PATCH 3/4] xfs_scrub: convert to per-ag inode bulkstat operations Darrick J. Wong @ 2019-09-12 23:55 ` Dave Chinner 0 siblings, 0 replies; 11+ messages in thread From: Dave Chinner @ 2019-09-12 23:55 UTC (permalink / raw) To: Darrick J. Wong; +Cc: sandeen, linux-xfs On Thu, Sep 05, 2019 at 08:36:03PM -0700, Darrick J. Wong wrote: > From: Darrick J. Wong <darrick.wong@oracle.com> > > Now that we're done reworking the xfrog bulkstat wrapper functions, we > can adapt xfs_scrub to use the per-ag iteration functionality. > > Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> > --- > scrub/fscounters.c | 22 ++++++---------------- > scrub/inodes.c | 20 ++++++-------------- > 2 files changed, 12 insertions(+), 30 deletions(-) Nice simplification. Reviewed-by: Dave Chinner <dchinner@redhat.com> -- Dave Chinner david@fromorbit.com ^ permalink raw reply [flat|nested] 11+ messages in thread
* [PATCH 4/4] xfs_scrub: batch inumbers calls during fscounters calculation 2019-09-06 3:35 [PATCH 0/4] xfsprogs: bulkstat v5 Darrick J. Wong ` (2 preceding siblings ...) 2019-09-06 3:36 ` [PATCH 3/4] xfs_scrub: convert to per-ag inode bulkstat operations Darrick J. Wong @ 2019-09-06 3:36 ` Darrick J. Wong 2019-09-12 23:56 ` Dave Chinner 3 siblings, 1 reply; 11+ messages in thread From: Darrick J. Wong @ 2019-09-06 3:36 UTC (permalink / raw) To: sandeen, darrick.wong; +Cc: linux-xfs From: Darrick J. Wong <darrick.wong@oracle.com> Improve the efficiency of the phase 7 inode counts by batching requests, now that we have per-AG inumbers wrappers. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> --- scrub/fscounters.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) diff --git a/scrub/fscounters.c b/scrub/fscounters.c index a2cf8171..ad467e0c 100644 --- a/scrub/fscounters.c +++ b/scrub/fscounters.c @@ -43,9 +43,10 @@ xfs_count_inodes_ag( { struct xfs_inumbers_req *ireq; uint64_t nr = 0; + unsigned int i; int error; - ireq = xfrog_inumbers_alloc_req(1, 0); + ireq = xfrog_inumbers_alloc_req(64, 0); if (!ireq) { str_info(ctx, descr, _("Insufficient memory; giving up.")); return false; @@ -55,7 +56,8 @@ xfs_count_inodes_ag( while (!(error = xfrog_inumbers(&ctx->mnt, ireq))) { if (ireq->hdr.ocount == 0) break; - nr += ireq->inumbers[0].xi_alloccount; + for (i = 0; i < ireq->hdr.ocount; i++) + nr += ireq->inumbers[i].xi_alloccount; } free(ireq); ^ permalink raw reply related [flat|nested] 11+ messages in thread
* Re: [PATCH 4/4] xfs_scrub: batch inumbers calls during fscounters calculation 2019-09-06 3:36 ` [PATCH 4/4] xfs_scrub: batch inumbers calls during fscounters calculation Darrick J. Wong @ 2019-09-12 23:56 ` Dave Chinner 0 siblings, 0 replies; 11+ messages in thread From: Dave Chinner @ 2019-09-12 23:56 UTC (permalink / raw) To: Darrick J. Wong; +Cc: sandeen, linux-xfs On Thu, Sep 05, 2019 at 08:36:11PM -0700, Darrick J. Wong wrote: > From: Darrick J. Wong <darrick.wong@oracle.com> > > Improve the efficiency of the phase 7 inode counts by batching requests, > now that we have per-AG inumbers wrappers. > > Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Makes sense. Reviewed-by: Dave Chinner <dchinner@redhat.com> -- Dave Chinner david@fromorbit.com ^ permalink raw reply [flat|nested] 11+ messages in thread
* [PATCH 0/4] xfsprogs: bulkstat v5 @ 2019-09-25 21:32 Darrick J. Wong 2019-09-25 21:33 ` [PATCH 2/4] xfs_spaceman: remove open-coded per-ag bulkstat Darrick J. Wong 0 siblings, 1 reply; 11+ messages in thread From: Darrick J. Wong @ 2019-09-25 21:32 UTC (permalink / raw) To: sandeen, darrick.wong; +Cc: linux-xfs Hi all, In this series, we do the heavier lifting to support the new v5 bulkstat code. First, we add an xfs_io command so that we can actually test the old and new interfaces. Then, we start converting spaceman and scrub to use the per-AG bulkstat functions so that they can shed a lot of duplicated geometry calculation code. Finally, we speed up scrub phase 7 by fixing a minor thinko. If you're going to start using this mess, you probably ought to just pull from my git trees, which are linked below. This is an extraordinary way to destroy everything. Enjoy! Comments and questions are, as always, welcome. --D xfsprogs git tree: https://git.kernel.org/cgit/linux/kernel/git/djwong/xfsprogs-dev.git/log/?h=bulkstat-v5-factoring ^ permalink raw reply [flat|nested] 11+ messages in thread
* [PATCH 2/4] xfs_spaceman: remove open-coded per-ag bulkstat 2019-09-25 21:32 [PATCH 0/4] xfsprogs: bulkstat v5 Darrick J. Wong @ 2019-09-25 21:33 ` Darrick J. Wong 0 siblings, 0 replies; 11+ messages in thread From: Darrick J. Wong @ 2019-09-25 21:33 UTC (permalink / raw) To: sandeen, darrick.wong; +Cc: linux-xfs, Dave Chinner From: Darrick J. Wong <darrick.wong@oracle.com> Now that xfrog_bulkstat supports per-AG bulkstat, we can get rid of the open coded one in spaceman. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Dave Chinner <dchinner@redhat.com> --- spaceman/health.c | 16 +++++----------- 1 file changed, 5 insertions(+), 11 deletions(-) diff --git a/spaceman/health.c b/spaceman/health.c index b195a229..8fd985a2 100644 --- a/spaceman/health.c +++ b/spaceman/health.c @@ -268,30 +268,24 @@ report_bulkstat_health( { struct xfs_bulkstat_req *breq; char descr[256]; - uint64_t startino = 0; - uint64_t lastino = -1ULL; uint32_t i; int error; - if (agno != NULLAGNUMBER) { - startino = cvt_agino_to_ino(&file->xfd, agno, 0); - lastino = cvt_agino_to_ino(&file->xfd, agno + 1, 0) - 1; - } - - breq = xfrog_bulkstat_alloc_req(BULKSTAT_NR, startino); + breq = xfrog_bulkstat_alloc_req(BULKSTAT_NR, 0); if (!breq) { perror("bulk alloc req"); exitcode = 1; return 1; } + if (agno != NULLAGNUMBER) + xfrog_bulkstat_set_ag(breq, agno); + do { error = xfrog_bulkstat(&file->xfd, breq); if (error) break; for (i = 0; i < breq->hdr.ocount; i++) { - if (breq->bulkstat[i].bs_ino > lastino) - goto out; snprintf(descr, sizeof(descr) - 1, _("inode %"PRIu64), breq->bulkstat[i].bs_ino); report_sick(descr, inode_flags, @@ -304,7 +298,7 @@ report_bulkstat_health( errno = error; perror("bulkstat"); } -out: + free(breq); return error; } ^ permalink raw reply related [flat|nested] 11+ messages in thread
* [PATCH 0/4] xfsprogs: bulkstat v5 @ 2019-08-26 21:27 Darrick J. Wong 2019-08-26 21:28 ` [PATCH 2/4] xfs_spaceman: remove open-coded per-ag bulkstat Darrick J. Wong 0 siblings, 1 reply; 11+ messages in thread From: Darrick J. Wong @ 2019-08-26 21:27 UTC (permalink / raw) To: sandeen, darrick.wong; +Cc: linux-xfs Hi all, In this series, we do the heavier lifting to support the new v5 bulkstat code. First, we add an xfs_io command so that we can actually test the old and new interfaces. Then, we start converting spaceman and scrub to use the per-AG bulkstat functions so that they can shed a lot of duplicated geometry calculation code. Finally, we speed up scrub phase 7 by fixing a minor thinko. If you're going to start using this mess, you probably ought to just pull from my git trees, which are linked below. This is an extraordinary way to destroy everything. Enjoy! Comments and questions are, as always, welcome. --D xfsprogs git tree: https://git.kernel.org/cgit/linux/kernel/git/djwong/xfsprogs-dev.git/log/?h=bulkstat-v5-factoring ^ permalink raw reply [flat|nested] 11+ messages in thread
* [PATCH 2/4] xfs_spaceman: remove open-coded per-ag bulkstat 2019-08-26 21:27 [PATCH 0/4] xfsprogs: bulkstat v5 Darrick J. Wong @ 2019-08-26 21:28 ` Darrick J. Wong 0 siblings, 0 replies; 11+ messages in thread From: Darrick J. Wong @ 2019-08-26 21:28 UTC (permalink / raw) To: sandeen, darrick.wong; +Cc: linux-xfs From: Darrick J. Wong <darrick.wong@oracle.com> Now that xfrog_bulkstat supports per-AG bulkstat, we can get rid of the open coded one in spaceman. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> --- spaceman/health.c | 16 +++++----------- 1 file changed, 5 insertions(+), 11 deletions(-) diff --git a/spaceman/health.c b/spaceman/health.c index ff03d074..43b99044 100644 --- a/spaceman/health.c +++ b/spaceman/health.c @@ -265,28 +265,22 @@ report_bulkstat_health( { struct xfs_bulkstat_req *breq; char descr[256]; - uint64_t startino = 0; - uint64_t lastino = -1ULL; uint32_t i; int error; - if (agno != NULLAGNUMBER) { - startino = xfrog_agino_to_ino(&file->xfd, agno, 0); - lastino = xfrog_agino_to_ino(&file->xfd, agno + 1, 0) - 1; - } - - breq = xfrog_bulkstat_alloc_req(128, startino); + breq = xfrog_bulkstat_alloc_req(128, 0); if (!breq) { perror("bulk alloc req"); exitcode = 1; return 1; } + if (agno != NULLAGNUMBER) + xfrog_bulkstat_set_ag(breq, agno); + while ((error = xfrog_bulkstat(&file->xfd, breq) == 0) && breq->hdr.ocount > 0) { for (i = 0; i < breq->hdr.ocount; i++) { - if (breq->bulkstat[i].bs_ino > lastino) - goto out; snprintf(descr, sizeof(descr) - 1, _("inode %"PRIu64), breq->bulkstat[i].bs_ino); report_sick(descr, inode_flags, @@ -294,7 +288,7 @@ report_bulkstat_health( breq->bulkstat[i].bs_checked); } } -out: + free(breq); return error; } ^ permalink raw reply related [flat|nested] 11+ messages in thread
end of thread, other threads:[~2019-09-25 21:33 UTC | newest] Thread overview: 11+ messages (download: mbox.gz / follow: Atom feed) -- links below jump to the message on this page -- 2019-09-06 3:35 [PATCH 0/4] xfsprogs: bulkstat v5 Darrick J. Wong 2019-09-06 3:35 ` [PATCH 1/4] xfs_io: add a bulkstat command Darrick J. Wong 2019-09-12 23:51 ` Dave Chinner 2019-09-06 3:35 ` [PATCH 2/4] xfs_spaceman: remove open-coded per-ag bulkstat Darrick J. Wong 2019-09-12 23:52 ` Dave Chinner 2019-09-06 3:36 ` [PATCH 3/4] xfs_scrub: convert to per-ag inode bulkstat operations Darrick J. Wong 2019-09-12 23:55 ` Dave Chinner 2019-09-06 3:36 ` [PATCH 4/4] xfs_scrub: batch inumbers calls during fscounters calculation Darrick J. Wong 2019-09-12 23:56 ` Dave Chinner -- strict thread matches above, loose matches on Subject: below -- 2019-09-25 21:32 [PATCH 0/4] xfsprogs: bulkstat v5 Darrick J. Wong 2019-09-25 21:33 ` [PATCH 2/4] xfs_spaceman: remove open-coded per-ag bulkstat Darrick J. Wong 2019-08-26 21:27 [PATCH 0/4] xfsprogs: bulkstat v5 Darrick J. Wong 2019-08-26 21:28 ` [PATCH 2/4] xfs_spaceman: remove open-coded per-ag bulkstat Darrick J. Wong
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for NNTP newsgroup(s).