| Message ID | 3ebe2d29-7943-b0a2-db5c-196610537bca@sandeen.net (mailing list archive) |
|---|---|
| State | Superseded |
| Headers | show |
| Series | [V2] libxfs: use FALLOC_FL_ZERO_RANGE in libxfs_device_zero | expand |
On Thu, Feb 13, 2020 at 07:05:50PM -0600, Eric Sandeen wrote: > I had a request from someone who cared about mkfs speed(!) > over a slower network block device to look into using faster > zeroing methods, particularly for the log, during mkfs. > > Using FALLOC_FL_ZERO_RANGE is faster in this case than writing > a bunch of zeros across a wire. > > Signed-off-by: Eric Sandeen <sandeen@redhat.com> > --- > > V2: Clean up all the nasty stuff I'd flung out there as a wild first > cut, thanks Dave. > > diff --git a/include/builddefs.in b/include/builddefs.in > index 4700b527..1dd27f76 100644 > --- a/include/builddefs.in > +++ b/include/builddefs.in > @@ -144,6 +144,9 @@ endif > ifeq ($(HAVE_GETFSMAP),yes) > PCFLAGS+= -DHAVE_GETFSMAP > endif > +ifeq ($(HAVE_FALLOCATE),yes) > +PCFLAGS += -DHAVE_FALLOCATE > +endif > > LIBICU_LIBS = @libicu_LIBS@ > LIBICU_CFLAGS = @libicu_CFLAGS@ > diff --git a/include/linux.h b/include/linux.h > index 8f3c32b0..8d5c4584 100644 > --- a/include/linux.h > +++ b/include/linux.h > @@ -20,6 +20,10 @@ > #include <stdio.h> > #include <asm/types.h> > #include <mntent.h> > +#include <fcntl.h> > +#if defined(HAVE_FALLOCATE) > +#include <linux/falloc.h> > +#endif > #ifdef OVERRIDE_SYSTEM_FSXATTR > # define fsxattr sys_fsxattr > #endif > @@ -164,6 +168,24 @@ static inline void platform_mntent_close(struct mntent_cursor * cursor) > endmntent(cursor->mtabp); > } > > +#if defined(FALLOC_FL_ZERO_RANGE) > +static inline int > +platform_zero_range( > + int fd, > + xfs_off_t start, > + size_t len) > +{ > + int ret; > + > + ret = fallocate(fd, FALLOC_FL_ZERO_RANGE, start, len); > + if (!ret) > + return 0; > + return -errno; > +} > +#else > +#define platform_zero_range(fd, s, l) (-EOPNOTSUPP) > +#endif That all looks good. :) > /* > * Check whether we have to define FS_IOC_FS[GS]ETXATTR ourselves. These > * are a copy of the definitions moved to linux/uapi/fs.h in the 4.5 kernel, > diff --git a/libxfs/rdwr.c b/libxfs/rdwr.c > index 0d9d7202..2f6a3eb3 100644 > --- a/libxfs/rdwr.c > +++ b/libxfs/rdwr.c > @@ -60,9 +60,19 @@ int > libxfs_device_zero(struct xfs_buftarg *btp, xfs_daddr_t start, uint len) > { > xfs_off_t start_offset, end_offset, offset; > - ssize_t zsize, bytes; > + ssize_t zsize, bytes, len_bytes; len_bytes should be size_t, right? > char *z; > - int fd; > + int error, fd; > + > + fd = libxfs_device_to_fd(btp->dev); > + start_offset = LIBXFS_BBTOOFF64(start); > + end_offset = LIBXFS_BBTOOFF64(start + len) - start_offset; > + > + /* try to use special zeroing methods, fall back to writes if needed */ > + len_bytes = LIBXFS_BBTOOFF64(len); > + error = platform_zero_range(fd, start_offset, len_bytes); This is a bit ... convoluted, and doesn't end_offset = len_bytes? i.e. start_offset = start << BBSHIFT len_bytes = len << BBSHIFT end_offset = (start + len) << BBSHIFT - start_offset = (start << BBSHIFT) + (len << BBSHIFT) - start_offset = start_offset + len_bytes - start_offset = len_bytes Cheers, Dave.
On 2/13/20 7:34 PM, Dave Chinner wrote: > On Thu, Feb 13, 2020 at 07:05:50PM -0600, Eric Sandeen wrote: ... >> /* >> * Check whether we have to define FS_IOC_FS[GS]ETXATTR ourselves. These >> * are a copy of the definitions moved to linux/uapi/fs.h in the 4.5 kernel, >> diff --git a/libxfs/rdwr.c b/libxfs/rdwr.c >> index 0d9d7202..2f6a3eb3 100644 >> --- a/libxfs/rdwr.c >> +++ b/libxfs/rdwr.c >> @@ -60,9 +60,19 @@ int >> libxfs_device_zero(struct xfs_buftarg *btp, xfs_daddr_t start, uint len) >> { >> xfs_off_t start_offset, end_offset, offset; >> - ssize_t zsize, bytes; >> + ssize_t zsize, bytes, len_bytes; > > len_bytes should be size_t, right? They probably all should be, TBH... >> char *z; >> - int fd; >> + int error, fd; >> + >> + fd = libxfs_device_to_fd(btp->dev); >> + start_offset = LIBXFS_BBTOOFF64(start); >> + end_offset = LIBXFS_BBTOOFF64(start + len) - start_offset; >> + >> + /* try to use special zeroing methods, fall back to writes if needed */ >> + len_bytes = LIBXFS_BBTOOFF64(len); >> + error = platform_zero_range(fd, start_offset, len_bytes); > > This is a bit ... convoluted, and doesn't end_offset = len_bytes? > i.e. > > start_offset = start << BBSHIFT > len_bytes = len << BBSHIFT > end_offset = (start + len) << BBSHIFT - start_offset > = (start << BBSHIFT) + (len << BBSHIFT) - start_offset > = start_offset + len_bytes - start_offset > = len_bytes oh, yikes. Good catch. Before, it was used as end_offset = LIBXFS_BBTOOFF64(start + len) - start_offset; for (offset = 0; offset < end_offset; ) { bytes = min((ssize_t)(end_offset - offset), zsize); if ((bytes = write(fd, z, bytes)) < 0) { "end_offset" was not the ending offset of the whole range... that's what I get for inferring too much from a variable name w/o looking at what it actually /does/ I'll sort that out, thanks. -Eric > Cheers, > > Dave. >
diff --git a/include/builddefs.in b/include/builddefs.in index 4700b527..1dd27f76 100644 --- a/include/builddefs.in +++ b/include/builddefs.in @@ -144,6 +144,9 @@ endif ifeq ($(HAVE_GETFSMAP),yes) PCFLAGS+= -DHAVE_GETFSMAP endif +ifeq ($(HAVE_FALLOCATE),yes) +PCFLAGS += -DHAVE_FALLOCATE +endif LIBICU_LIBS = @libicu_LIBS@ LIBICU_CFLAGS = @libicu_CFLAGS@ diff --git a/include/linux.h b/include/linux.h index 8f3c32b0..8d5c4584 100644 --- a/include/linux.h +++ b/include/linux.h @@ -20,6 +20,10 @@ #include <stdio.h> #include <asm/types.h> #include <mntent.h> +#include <fcntl.h> +#if defined(HAVE_FALLOCATE) +#include <linux/falloc.h> +#endif #ifdef OVERRIDE_SYSTEM_FSXATTR # define fsxattr sys_fsxattr #endif @@ -164,6 +168,24 @@ static inline void platform_mntent_close(struct mntent_cursor * cursor) endmntent(cursor->mtabp); } +#if defined(FALLOC_FL_ZERO_RANGE) +static inline int +platform_zero_range( + int fd, + xfs_off_t start, + size_t len) +{ + int ret; + + ret = fallocate(fd, FALLOC_FL_ZERO_RANGE, start, len); + if (!ret) + return 0; + return -errno; +} +#else +#define platform_zero_range(fd, s, l) (-EOPNOTSUPP) +#endif + /* * Check whether we have to define FS_IOC_FS[GS]ETXATTR ourselves. These * are a copy of the definitions moved to linux/uapi/fs.h in the 4.5 kernel, diff --git a/libxfs/rdwr.c b/libxfs/rdwr.c index 0d9d7202..2f6a3eb3 100644 --- a/libxfs/rdwr.c +++ b/libxfs/rdwr.c @@ -60,9 +60,19 @@ int libxfs_device_zero(struct xfs_buftarg *btp, xfs_daddr_t start, uint len) { xfs_off_t start_offset, end_offset, offset; - ssize_t zsize, bytes; + ssize_t zsize, bytes, len_bytes; char *z; - int fd; + int error, fd; + + fd = libxfs_device_to_fd(btp->dev); + start_offset = LIBXFS_BBTOOFF64(start); + end_offset = LIBXFS_BBTOOFF64(start + len) - start_offset; + + /* try to use special zeroing methods, fall back to writes if needed */ + len_bytes = LIBXFS_BBTOOFF64(len); + error = platform_zero_range(fd, start_offset, len_bytes); + if (!error) + return 0; zsize = min(BDSTRAT_SIZE, BBTOB(len)); if ((z = memalign(libxfs_device_alignment(), zsize)) == NULL) { @@ -73,9 +83,6 @@ libxfs_device_zero(struct xfs_buftarg *btp, xfs_daddr_t start, uint len) } memset(z, 0, zsize); - fd = libxfs_device_to_fd(btp->dev); - start_offset = LIBXFS_BBTOOFF64(start); - if ((lseek(fd, start_offset, SEEK_SET)) < 0) { fprintf(stderr, _("%s: %s seek to offset %llu failed: %s\n"), progname, __FUNCTION__, @@ -83,7 +90,6 @@ libxfs_device_zero(struct xfs_buftarg *btp, xfs_daddr_t start, uint len) exit(1); } - end_offset = LIBXFS_BBTOOFF64(start + len) - start_offset; for (offset = 0; offset < end_offset; ) { bytes = min((ssize_t)(end_offset - offset), zsize); if ((bytes = write(fd, z, bytes)) < 0) {
I had a request from someone who cared about mkfs speed(!) over a slower network block device to look into using faster zeroing methods, particularly for the log, during mkfs. Using FALLOC_FL_ZERO_RANGE is faster in this case than writing a bunch of zeros across a wire. Signed-off-by: Eric Sandeen <sandeen@redhat.com> --- V2: Clean up all the nasty stuff I'd flung out there as a wild first cut, thanks Dave.