OSDN Git Service
Jim Meyering [Mon, 2 Nov 2009 18:44:40 +0000 (19:44 +0100)]
tests: serialize tests that use the scsi_debug module
* tests/t-local.sh (scsi_debug_lock_file_): New global.
(scsi_debug_acquire_lock_): New function.
(scsi_debug_setup_): Call it, to acquire the lock.
(scsi_debug_cleanup_): Release the lock.
Jim Meyering [Mon, 2 Nov 2009 14:34:34 +0000 (15:34 +0100)]
tests: test the new alignment functions
* tests/t9020-alignment.sh: New test.
* tests/Makefile.am (TESTS): Add it to the list.
Jim Meyering [Mon, 2 Nov 2009 14:33:15 +0000 (15:33 +0100)]
tests: factor out some more
* tests/t-local.sh (require_scsi_debug_module_): New function.
* tests/t9010-big-sector.sh: Use it.
Jim Meyering [Mon, 2 Nov 2009 11:12:24 +0000 (12:12 +0100)]
tests: perform scsi_debug cleanup automatically...
...rather than requiring each test to do it.
* tests/t-lib.sh ($cleanup_eval): New variable. Initialize.
(remove_tmp_): Use it.
* tests/t-local.sh: Record whether modprobe succeeded.
(cleanup_eval_): Append an invocation of scsi_debug_cleanup_
so that it is always run.
(scsi_debug_cleanup_): Run rmmod only if the modprobe succeeded.
* tests/t9010-big-sector.sh (cleanup_): Remove.
Jim Meyering [Mon, 2 Nov 2009 10:35:02 +0000 (11:35 +0100)]
tests: move scsi_debug framework into shared script
* tests/t-local.sh (-scsi_debug_cleanup_, wait_for_dev_to_appear_):
(print_sd_names_, scsi_debug_setup_): New functions, factored out of ...
* tests/t9010-big-sector.sh: ...this file.
Jim Meyering [Mon, 2 Nov 2009 10:21:06 +0000 (11:21 +0100)]
tests: document and begin to factor out scsi_debug framework
* tests/t9010-big-sector.sh: Add comments, fix a bug in
"comm" usage, prepare to factor out scsi_debug framework.
Jim Meyering [Sat, 31 Oct 2009 23:56:30 +0000 (00:56 +0100)]
tests: add a helper program to print alignment info for a device
* tests/print-align.c: New file.
* tests/Makefile.am (check_PROGRAMS): Define.
(AM_CPPFLAGS, LDADD): Define.
(AM_CFLAGS): Enable warnings here, too, by adding
$(WARN_CFLAGS) and $(WERROR_CFLAGS).
Hans de Goede [Fri, 30 Oct 2009 11:59:05 +0000 (12:59 +0100)]
Implement get_partition_alignment DiskOp for all labels which need it
Implement get_partition_alignment DiskOp for all labels which require
an alignment with a grain_size != 1.
Hans de Goede [Fri, 30 Oct 2009 11:59:04 +0000 (12:59 +0100)]
libparted: add ped_disk_get_partition_alignment() function
This functions get the alignment needed for partition boundaries on this disk.
The returned alignment describes the alignment for the start sector of the
partition, for all disklabel types which require alignment, except Sun
disklabels, the end sector must be aligned too. To get the end sector
alignment decrease the PedAlignment offset by 1.
Hans de Goede [Fri, 30 Oct 2009 11:59:03 +0000 (12:59 +0100)]
libparted: add ped_device_get_xxx_aligned_constraint() functions
Add ped_device_get_minimal_aligned_constraint() and
ped_device_get_optimal_aligned_constraint() functions to libparted.
These functions can be used to get a device constraint taking alignment into
account.
Hans de Goede [Fri, 30 Oct 2009 11:59:02 +0000 (12:59 +0100)]
ped_device_get_constraint: return constraint without alignment
In libparted-1.9 ped_device_get_constraint() would always return a constraint
without any start / end aligning (grain_size == 1). Now that we are starting to
support disks with a physical sector size other then 512, the old
ped_device_get_constraint() code may actually start returning constrains
with alignment info with a grain_size != 1.
Since this is a behavior change and since the old code does not take a possible
alignment info into account, this patch removes the aligning code from
ped_device_get_constraint(), so that its behavior is unchanged.
This patch also updates the functions doxygen documentation to reflect what
it has actually done all these years.
Hans de Goede [Fri, 30 Oct 2009 11:59:01 +0000 (12:59 +0100)]
linux: add get_xxx_alignment() functions
* libparted/arch/linux.c (get_minimum_alignment): New function.
(get_optimum_alignment): New function.
Hans de Goede [Fri, 30 Oct 2009 11:59:00 +0000 (12:59 +0100)]
linux: save blkid topology in arch_specific data for later use
This is a preparation patch for adding get_minimum_alignment
and get_optimum_alignment functions to linux's _PedDeviceArchOps.
* libparted/arch/linux.h (struct_LinuxSpecific) [probe, topology]:
Add members.
* libparted/arch/linux.c (linux_new): initialize
arch_specific->probe and arch_specific->topology to NULL,
(linux_destroy): Free arch_specific->probe.
(get_minimum_io_size): Remove function.
(get_blkid_topology): New function.
(_device_set_sector_size): Call get_blkid_topology() instead of
get_minimum_io_size() and retrieve minimum io size from the cached
topology.
Hans de Goede [Fri, 30 Oct 2009 11:58:59 +0000 (12:58 +0100)]
libparted: add ped_device_get_xxx_alignment() functions
Add ped_device_get_minimum_alignment() and ped_device_get_optimum_alignment()
functions to libparted.
Note this is a resent of my previous patchset with a number of typos corrected:
aligment -> alignment
minimal_alignment -> minimum_aligment
optimal_alignment -> optimum_aligment
Sorry for the noise.
Jim Meyering [Thu, 29 Oct 2009 12:29:26 +0000 (13:29 +0100)]
tests: protect t7000 from the environment
* tests/t7000-scripting.sh: Initialize fail=0, so that an
environment setting cannot perturb the test.
Jim Meyering [Thu, 29 Oct 2009 11:34:08 +0000 (12:34 +0100)]
tests: convert to new testing framework
* tests/t0200-gpt.sh: Convert to new testing framework.
* tests/t0201-gpt.sh: Likewise.
* tests/t0300-dos-on-gpt.sh: Likewise.
* tests/t0100-print.sh: Likewise.
* tests/t0250-gpt.sh: Likewise.
Jim Meyering [Thu, 29 Oct 2009 11:22:25 +0000 (12:22 +0100)]
dvh: document an as-yet-unplugged (but not serious) leak
* libparted/disk.c (ped_disk_add_partition): List some of the
things I've tried to plug a leak exposed by a root-only test
on dvh's partition-creation error path. This leak is enough
of a corner case that I may even leave it for now, and add a
suppression for it, so that we're not distracted by it.
Jim Meyering [Wed, 28 Oct 2009 16:25:46 +0000 (17:25 +0100)]
dvh: plug three constraint-sized leaks
* libparted/labels/dvh.c (dvh_read, dvh_alloc_metadata):
Free constraint upon failure, too.
Jim Meyering [Wed, 28 Oct 2009 15:40:07 +0000 (16:40 +0100)]
ui: plug multiple command_line_get_word leaks
* parted/parted.c (do_mkpart, do_print):
Jim Meyering [Wed, 28 Oct 2009 15:37:07 +0000 (16:37 +0100)]
check: plug an even smaller (formatted "unit") leak
* libparted/disk.c (ped_disk_check): Free fs_size also upon success.
Jim Meyering [Wed, 28 Oct 2009 15:19:43 +0000 (16:19 +0100)]
check: plug a geometry-sized leak
* libparted/disk.c (ped_disk_check): Destroy "geom", when done with it.
Jim Meyering [Wed, 28 Oct 2009 15:03:11 +0000 (16:03 +0100)]
loop: don't leak a sector in loop_clobber
* libparted/labels/loop.c (loop_clobber): Replace leaky open-coded
version with a use of Use ptt_clear_sectors.
Jim Meyering [Fri, 18 Sep 2009 06:40:54 +0000 (08:40 +0200)]
tests: t7000: remove mkpartfs tests
This test tested both mkpart and mkpartfs.
* tests/t7000-scripting.sh: Remove the mkpartfs tests.
Jim Meyering [Thu, 17 Sep 2009 18:12:50 +0000 (20:12 +0200)]
tests: remove tests that are too FS-centric
* tests/t1000-mkpartfs.sh: Remove file.
* tests/t1500-small-ext2.sh: Likewise.
* tests/t2000-mkfs.sh: Likewise.
* tests/t3100-resize-ext2-partion.sh: Likewise.
* tests/t3000-constraints.sh: Likewise.
* tests/Makefile.am (TESTS): Remove them here, too.
Jim Meyering [Tue, 27 Oct 2009 22:10:35 +0000 (23:10 +0100)]
ui: plug a constraint leak in do_mkpart
* parted/parted.c (do_mkpart): Free constraint after using it to
add partition.
Jim Meyering [Tue, 27 Oct 2009 21:44:04 +0000 (22:44 +0100)]
gpt: plug two gpt-header leaks in gpt_read_headers
* libparted/labels/gpt.c (gpt_read_headers): Free locals pri and/or
bak if they are not valid.
Jim Meyering [Tue, 27 Oct 2009 21:38:02 +0000 (22:38 +0100)]
gpt: plug a 1-sector leak in gpt_read_headers
* libparted/labels/gpt.c (gpt_read_headers): Free a sector.
Jim Meyering [Tue, 27 Oct 2009 20:42:25 +0000 (21:42 +0100)]
ui: plug a leak (partition_print_flags)
* parted/parted.c (do_print): Free flags, after printing.
Jim Meyering [Tue, 27 Oct 2009 20:36:30 +0000 (21:36 +0100)]
libparted: plug an FS-alias-related leak
* libparted/filesys.c (ped_file_system_alias_unregister):
Free the alias we've just unregistered.
Jim Meyering [Mon, 26 Oct 2009 14:34:55 +0000 (15:34 +0100)]
doc: update NEWS
* NEWS (Bug fixes): Describe two.
Jim Meyering [Fri, 23 Oct 2009 14:32:20 +0000 (16:32 +0200)]
gpt: header validity: also consider whether a PE array crc matches
* libparted/labels/gpt.c (check_PE_array_CRC): New function.
(_header_is_valid): Use it to verify that partition entry array
checksums match, too.
Karel Zak noticed that this check was not being performed.
* tests/t0280-gpt-corrupt.sh: Adapt to changed diagnostic.
Jim Meyering [Fri, 23 Oct 2009 14:15:19 +0000 (16:15 +0200)]
gpt: tweak interfaces to pass disk, not disk->dev
* libparted/labels/gpt.c: required for upcoming CRC-checking function.
Jim Meyering [Fri, 23 Oct 2009 10:48:11 +0000 (12:48 +0200)]
gpt: don't ignore an improbable (1-in-2^32) failure
pth_crc32 mapped failure to a crc32 value of 0. If that happened
to match the stored value, there would be no indication that there
had been a malloc failure.
* libparted/labels/gpt.c (pth_crc32): Change interface so we can
distinguish failure from a CRC value of 0.
Update all callers.
(_generate_header): Now that we recognize pth_crc32 can fail,
we must admit that its void-returning caller _generate_header
may fail, too. Return "int" and adjust *its* callers.
Jim Meyering [Thu, 22 Oct 2009 20:14:03 +0000 (22:14 +0200)]
gpt: factor out code that is about to be reused
* libparted/labels/gpt.c: Include <errno.h>.
(gpt_read_PE_array): New function, ...
(gpt_read): ...factored out of this one. Use it.
Jim Meyering [Thu, 22 Oct 2009 19:46:00 +0000 (21:46 +0200)]
gpt: don't dereference NULL upon OOM (pth_get_raw callers)
Handle each case in which pth_get_raw may return NULL.
* libparted/labels/gpt.c (pth_get_raw): Upon OOM error, return 0
rather than dereferencing NULL.
(gpt_write): Don't dereference NULL upon OOM error.
(gpt_read): Yet another.
Jim Meyering [Thu, 22 Oct 2009 19:34:48 +0000 (21:34 +0200)]
gpt: don't dereference NULL upon OOM (in pth_get_raw)
* libparted/labels/gpt.c (pth_get_raw): Don't dereference NULL
on out of memory error.
Jim Meyering [Thu, 22 Oct 2009 19:26:26 +0000 (21:26 +0200)]
gpt: remove dead store
* libparted/labels/gpt.c (pth_crc32): Remove useless initialization.
Jim Meyering [Thu, 22 Oct 2009 18:55:39 +0000 (20:55 +0200)]
gpt: make an internal function "const"-correct
* libparted/labels/gpt.c (_parse_header): Make "gpt" const.
Jim Meyering [Thu, 22 Oct 2009 18:18:48 +0000 (20:18 +0200)]
tests: convert t0280 to new test style.
* tests/t0280-gpt-corrupt.sh: Convert to new test style.
* tests/t-lib.sh: Source t-local.sh.
* tests/t-local.sh: New file.
Jim Meyering [Thu, 22 Oct 2009 16:08:57 +0000 (18:08 +0200)]
gpt: clean up
* libparted/labels/gpt.c (gpt_probe): Remove now-useless test before
pth_free.
(gpt_read): Use ptt_clear_sectors in place of open-coded equivalent.
* cfg.mk (useless_free_options): Add --name=pth_free.
Jim Meyering [Thu, 22 Oct 2009 09:35:44 +0000 (11:35 +0200)]
gpt: make its header check a tiny bit more strict
* libparted/labels/gpt.c (_header_is_valid): Declare a header
to be invalid if its AlternateLBA is the same as MyLBA.
2009-10-21 Jim Meyering <meyering@redhat.com>
complete most of gpt-CRC-related rewrite
* libparted/labels/gpt.c (_read_header): Remove now-unused function.
Jim Meyering [Wed, 21 Oct 2009 16:21:07 +0000 (18:21 +0200)]
t0280-gpt-corrupt.sh: adapt for new output, improve test
Jim Meyering [Wed, 21 Oct 2009 11:33:20 +0000 (13:33 +0200)]
gpt: rewrite GPT header-reading code
* libparted/labels/gpt.c (read_header): Remove now-unused function.
(gpt_read_headers): New function.
(gpt_read): Rewrite to use gpt_read_headers.
Jim Meyering [Mon, 19 Oct 2009 18:14:44 +0000 (20:14 +0200)]
gpt: require that the backup table's AlternateLBA is 1
* libparted/labels/gpt.c (_header_is_valid): When checking
the backup header, ensure that its AlternateLBA is 1.
Jim Meyering [Mon, 19 Oct 2009 18:12:41 +0000 (20:12 +0200)]
tests: gpt: test for detection of corrupt backup partition header
* tests/t0280-gpt-corrupt.sh: Corrupt the backup partition header's
MyLBA, and require that an attempt to print the table fails.
Jim Meyering [Mon, 19 Oct 2009 16:55:06 +0000 (18:55 +0200)]
tests: hook up new GPT-crc-corrupting script
* tests/Makefile.am (TESTS): Add t0280-gpt-corrupt.sh.
* libparted/labels/gpt.c (gpt_read): Change diagnostic to say that
the PTE CRC mismatch is in the primary table.
* tests/t0280-gpt-corrupt.sh: Tweak diagnostic to match
what the code now prints.
Jim Meyering [Mon, 19 Oct 2009 16:47:00 +0000 (18:47 +0200)]
gpt: also verify MyLBA when checking table validity
* libparted/labels/gpt.c (_header_is_valid): Add parameter, my_lba.
(_read_header): Update caller.
(_parse_header): Comment out this use of _header_is_valid
in an assertion.
Jim Meyering [Mon, 19 Oct 2009 16:03:46 +0000 (18:03 +0200)]
gpt: clean up formatting of static initializers
* libparted/labels/gpt.c (NULL_IF_DISCOVER_ONLY): Define.
(gpt_disk_ops): Use it to avoid in-decl #if directives.
Jim Meyering [Mon, 19 Oct 2009 15:34:19 +0000 (17:34 +0200)]
gpt.c: filter through indent --no-tabs...
and manual fix-up to shorten long lines.
Also, manually format the static initializers in e.g., gpt_disk_ops,
since GNU indent botches that.
Jim Meyering [Mon, 19 Oct 2009 14:56:21 +0000 (16:56 +0200)]
gpt: add a comment
* libparted/labels/gpt.c (_header_is_valid): Add a comment to explain
the new partition entry size limitation of UINT32_MAX/16.
Jim Meyering [Mon, 19 Oct 2009 12:41:00 +0000 (14:41 +0200)]
build: update gnulib submodule to latest
Jim Meyering [Fri, 16 Oct 2009 17:14:38 +0000 (19:14 +0200)]
tests: use scsi_debug to test-for 4k-sector-specific bug
* tests/t9010-big-sector.sh: New root-only test.
Requires 2.6.31, and the scsi_debug module, /sys/block/... support
and an upcoming version of libblkid.
* tests/Makefile.am (TESTS): Add t9010-big-sector.sh.
Thanks to Mike Snitzer for advice and for pointing me to similar
code he wrote to test LVM.
Jim Meyering [Mon, 19 Oct 2009 10:03:20 +0000 (12:03 +0200)]
tests: begin the transition to coreutils' testing framework
The existing testing framework makes it hard to diagnose test failures,
so I'm switching to one that works better for me, from coreutils.
* tests/t-lib.sh: New file. Mostly from coreutils/tests/test-lib.sh.
* tests/Makefile.am (TMPDIR): Update to pass more info to test scripts.
(EXTRA_DIST): Add t-lib.sh.
Jim Meyering [Thu, 15 Oct 2009 19:29:36 +0000 (21:29 +0200)]
linux: use libblkid to determine ->phys_sector_size
Before this change, creating a memory-mapped disk on a fedora-based
system running 2.6.31.1-56.fc12.x86_64 using this command:
modprobe scsi_debug dev_size_mb=1025 sector_size=4096
and then running "parted -s /dev/sdd mklabel gpt print"
would mistakenly print "Sector size (logical/physical): 4096B/512B"
The "512B" is what's wrong. It should be "4096B".
* configure.ac: Test for a new-enough blkid library.
* libparted/Makefile.am (libparted_la_LIBADD): Add $(LIB_BLKID).
* libparted/arch/linux.c (get_minimum_io_size): New function.
(_device_set_sector_size): Use it.
Jim Meyering [Wed, 14 Oct 2009 18:53:48 +0000 (20:53 +0200)]
libparted: weaken a diagnostic
* libparted/arch/linux.c (_kernel_reread_part_table): Give a more
accurate diagnostic. Reported by Curtis Gedak.
Jim Meyering [Wed, 14 Oct 2009 18:50:39 +0000 (20:50 +0200)]
gpt: verify CRC32 of partition table entry array, too
The partition table entry array's CRC was not being verified at all.
This adds an interim check. The full-blown change required to make
this implementation conform to the EFI spec will be more invasive.
* libparted/labels/gpt.c (gpt_read): Check CRC of PTE array, too.
* tests/t0280-gpt-corrupt.sh: New test for this.
Reported by Karel Zak.
Jim Meyering [Wed, 14 Oct 2009 10:34:56 +0000 (12:34 +0200)]
doc: mention the "bios_grub" flag
I don't like the idea of documenting this "flag setting" option,
but it's been in the code for long enough now, and in actual use,
that I feel obliged. In the longer term, we'll probably deprecate
it and instead add an interface to set a partition's type independent
of mkpart's "type" argument.
* doc/parted.texi (set): Mention bios_grub.
Jim Meyering [Wed, 14 Oct 2009 09:36:47 +0000 (11:36 +0200)]
tests: gpt: do not munge GPT data manually
Otherwise, an upcoming change would cause this test to fail.
* tests/t5000-tags.sh: Do not munge GPT data manually, since that
does not update embedded CRCs, and thus renders it invalid.
Set the attribute via "set 1 bios_grub on" instead.
Also, use "name1" as the partition name, rather than the
misleading "primary".
Adjust diagnostics: the changes we make are no longer "manual."
Jim Meyering [Thu, 8 Oct 2009 20:03:37 +0000 (22:03 +0200)]
libparted: fix big-endian bug in gpt_get_max_supported_partition_count
* libparted/labels/gpt.c (gpt_get_max_supported_partition_count):
Convert ->SizeOfPartitionEntry from little-endian to CPU endianness
here, too. This is along the same lines as commit
d6e9b3bb, but
with the difference that this code is not reachable via the
command-line interface.
Jim Meyering [Tue, 6 Oct 2009 18:38:19 +0000 (20:38 +0200)]
post-release administrivia
* NEWS: Add header line for next release.
* .prev-version: Record previous version.
* cfg.mk (old_NEWS_hash): Auto-update.
Jim Meyering [Tue, 6 Oct 2009 18:29:16 +0000 (20:29 +0200)]
version 2.0
* NEWS: Record release date.
Jim Meyering [Tue, 6 Oct 2009 18:10:00 +0000 (20:10 +0200)]
gpt: don't malfunction on big-endian systems
Numerous GPT tests would fail when run on e.g., big-endian PPC.
* libparted/labels/gpt.c (gpt_read): Now that we use the
SizeOfPartitionEntry member, be sure to convert from GPT's
little-endian on-disk format to to CPU endianness.
This bug was introduced via commit
14cce9b2, 2009-06-10, "gpt:
fix gpt_read to read all of the partition entries correctly".
* libparted/labels/gpt.c (_header_is_valid): Also convert it here.
Add a test to ensure that the partition entry size is no larger
than the slightly arbitrary UINT32_MAX/16.
* NEWS (Bug fixes): Mention it.
Jim Meyering [Tue, 6 Oct 2009 14:56:45 +0000 (16:56 +0200)]
build: update gnulib submodule to latest
Jim Meyering [Tue, 6 Oct 2009 14:39:13 +0000 (16:39 +0200)]
maint: add gnulib's gnu-web-doc-update module
* bootstrap.conf (gnulib_modules): Add gnu-web-doc-update
Jim Meyering [Tue, 6 Oct 2009 14:26:00 +0000 (16:26 +0200)]
doc: add pointer to NEWS in README
* README: Mention NEWS.
Jim Meyering [Tue, 6 Oct 2009 14:23:19 +0000 (16:23 +0200)]
doc: update README
* README (WARNING): Mention upcoming removal of FS-manipulation
capabilities and that we'll retain FAT/HFS resize capabilities.
Remove an obsolete URL. Do not encourage email to individuals.
Update info-viewing instructions, provide the on-line doc URL.
Jim Meyering [Sat, 19 Sep 2009 06:22:20 +0000 (08:22 +0200)]
tests: resize FAT and HFS file systems
Since it looks like we're stuck with keeping at least FAT and HFS
file-system resizing support in libparted, we have an obligation to
exercise at least some of that functionality. For now, I've limited
this test to shrinking well-aligned partitions by a well-behaved
amount. Curtis Gedak helped come up with a usable test.
* tests/t3000-resize-fs.sh: New file.
Use starting partition offset that is on a cylinder boundary,
and increase the partition size to be larger than 256 MiB.
When resizing, choose the new size to be exactly one cylinder
(8MiB) smaller than the original.
Hans de Goede [Tue, 6 Oct 2009 09:31:14 +0000 (11:31 +0200)]
ped_partition_busy: do not call the libparted exception handler
Modify libparted/arch/linux.c _partition_get_part_dev() to not call
_device_stat() but instead use stat directly, as _device_stat() calls
the libparted exception handler, and we don't want this here, the only
caller of _partition_get_part_dev() is _partition_is_mounted(), which
in turn only gets called by linux_partition_is_busy(), and we don't
want to "throw" exceptions from ped_partition_busy().
This issue was noticed in combination with pyparted as used by
anaconda, see: http://bugzilla.redhat.com/527035#c10
Jim Meyering [Thu, 1 Oct 2009 18:00:32 +0000 (20:00 +0200)]
doc: mention the GPT/msftres bug fix
* NEWS (Bug fixes): Mention it.
Jim Meyering [Thu, 1 Oct 2009 13:01:52 +0000 (15:01 +0200)]
parted: issue a warning upon any use of a file-system op
* parted/parted.c (issue_fs_op_warning): New function.
(do_cp, do_mkfs, do_mkpartfs, do_move, do_resize, do_check): Call it.
* tests/test-lib.sh: Export the envvar setting,
PARTED_SUPPRESS_FILE_SYSTEM_MANIPULATION_WARNING=1 to suppress
those warnings while running the few remaining FS-related tests.
Jim Meyering [Wed, 30 Sep 2009 09:42:32 +0000 (11:42 +0200)]
maint: avoid false-positive NULL-deref warning from clang
* parted/ui.c: Include <assert.h>.
(command_line_get_disk): Add an assertion that command_line_get_device
currently guarantees will always be true. This placates clang regarding
its sole NULL-deref warning.
(command_line_get_device): Move declarations "down" to first use.
Jim Meyering [Thu, 17 Sep 2009 12:15:01 +0000 (14:15 +0200)]
build: use more gnulib modules for better POSIX compliance
* bootstrap.conf (gnulib_modules): Add modules exposed via
make CFLAGS=-DGNULIB_POSIXCHECK 2>&1 \
|perl -lne '/.* use gnulib module (\S+).*/ and print $1' \
|sort |uniq -c|sort -nr
Add these: close fsync lseek mkstemp strdup unlink
Jim Meyering [Tue, 29 Sep 2009 20:02:56 +0000 (22:02 +0200)]
maint: function returning "int", did not return a value.
* parted/ui.c (init_readline): Return a value (always 0 for now).
* parted/parted.c (_init): Handle init_readline's return value.
Jim Meyering [Tue, 29 Sep 2009 19:49:55 +0000 (21:49 +0200)]
maint: remove more dead stores and fix a bug in do_mkpartfs
* libparted/fs/ext2/ext2_mkfs.c (ext2_mkfs_create_lost_and_found_inode):
Punt: suppress clang's (aka scan-build's) warning.
* parted/parted.c (snap_to_boundaries): Remove a dead store.
(do_mkpartfs): The apparent dead store to part_name exposed a real bug.
In fact, it highlighted that while do_mkpart uses its "part_name"
variable (and calls ped_partition_set_name), do_mkpartfs did not.
(snap_to_boundaries): Remove dead store.
* libparted/unit.c (parse_chs): Likewise, and a leak.
* libparted/fs/hfs/reloc_plus.c (hfsplus_cache_from_catalog): Likewise.
(hfsplus_cache_from_extent): Remove a dead store.
(hfsplus_cache_from_attributes): Likewise.
* libparted/fs/hfs/advfs_plus.c (hfsplus_btree_search): Likewise.
* libparted/fs/ext2/ext2_block_relocator.c (ext2_block_relocate_shrink):
Remove dead store and corresponding decl.
* parted/ui.h: Mark 2nd parameter as non-null.
Jim Meyering [Tue, 29 Sep 2009 19:29:00 +0000 (21:29 +0200)]
maint: remove dead store code and declarations
* libparted/labels/bsd.c (bsd_partition_set_flag): Remove dead store
and corresponding decl.
* libparted/arch/linux.c (init_ide): Likewise.
(_dm_maptype): Likewise.
(_mount_table_search): Likewise.
* libparted/fs/amiga/asfs.c (_asfs_probe): Likewise.
* libparted/labels/mac.c (mac_alloc_metadata): Likewise.
* libparted/labels/rdb.c (amiga_write): Likewise.
(amiga_write): Again.
* libparted/fs/amiga/apfs.c (_generic_apfs_probe): Likewise.
* libparted/fs/hfs/reloc_plus.c (hfsplus_effect_move_extent): Likewise.
* libparted/fs/hfs/reloc.c (hfs_effect_move_extent): Likewise.
* libparted/labels/dos.c (probe_partition_for_geom): Remove dead
store in "can't happen" case.
Jim Meyering [Wed, 30 Sep 2009 09:10:15 +0000 (11:10 +0200)]
make PED_ASSERT handling sane: abort on false condition
Upon a failed PED_ASSERT, Do not ask the interactive user if they want
to continue. This change will affect any code that expects to be able
to continue after a failed PED_ASSERT condition. However, such code is
so fundamentally broken that this change is required. If you require
to be able to continue after a false condition, then the code must
not use a macro named like PED_ASSERT. This change was motivated by
my desire to use the clang static analysis tool. Without this change,
there were over 300 mostly-false-positive reports. With it, just 31,
almost all legitimate.
* libparted/debug.c (ped_assert): Remove first parameter.
It is now tested via the macro.
Now, this function always aborts, which helps clang understand.
Do not ask the interactive user if s/he wants to continue.
* include/parted/debug.h (PED_ASSERT): Test condition here,
not in ped_assert.
* libparted/disk.c (ped_disk_remove_partition): Remove the label
that was used only upon failing PED_ASSERT. Now that PED_ASSERT
always aborts for a false condition, that would have been dead code.
(ped_disk_new_fresh): Likewise.
Jim Meyering [Thu, 1 Oct 2009 10:03:07 +0000 (12:03 +0200)]
build: avoid a shadowing warning
* libparted/fs/hfs/advfs_plus.c (hfsplus_get_empty_end): Rename local,
"link" to avoid shadowing the syscall.
* libparted/fs/hfs/advfs.c (hfs_get_empty_end): Likewise.
Jim Meyering [Fri, 25 Sep 2009 06:53:30 +0000 (08:53 +0200)]
fat-FS: do not modify fat file system support
Revert "Add fat support for sector_size > 512."
This reverts commit
14395c356664a3b88f3929a6a3ad69bca374f70d.
Jim Meyering [Fri, 25 Sep 2009 06:53:18 +0000 (08:53 +0200)]
Revert "Fix compile failures."
This reverts commit
7603e404ee60f2f892be380a23c390f4638f451f.
Jim Meyering [Tue, 29 Sep 2009 15:16:59 +0000 (17:16 +0200)]
doc: update NEWS
* NEWS: Mention >512-byte sector support.
Jim Meyering [Tue, 29 Sep 2009 14:58:41 +0000 (16:58 +0200)]
build: use newer bootstrap script
* bootstrap: Update from coreutils.
Jim Meyering [Tue, 29 Sep 2009 14:57:08 +0000 (16:57 +0200)]
build: update gnulib submodule to latest
Jim Meyering [Tue, 29 Sep 2009 14:55:37 +0000 (16:55 +0200)]
maint: update most copyright year lists to include 2009
I used this command:
make update-copyright \
UPDATE_COPYRIGHT_FORCE=1 UPDATE_COPYRIGHT_USE_INTERVALS=1
Jim Meyering [Tue, 29 Sep 2009 14:52:18 +0000 (16:52 +0200)]
tests: enable more syntax-checks
* cfg.mk: Enable a few more tests.
* include/parted/debug.h: Remove useless parens to avoid
triggering failure of the sc_useless_cpp_parens syntax-check.
* libparted/arch/linux.h: Likewise.
* libparted/libparted.c: Likewise.
Jim Meyering [Tue, 29 Sep 2009 14:46:16 +0000 (16:46 +0200)]
tests: enable a few syntax-checks
* cfg.mk (local-checks-to-skip): Remove a few useless exclusions
and enable a few syntax checks.
* libparted/arch/linux.c (linux_destroy): Avoid triggering the
sc_cast_of_argument_to_free check.
* parted/strlist.c (str_list_destroy_node): Likewise.
Jim Meyering [Thu, 24 Sep 2009 17:56:37 +0000 (19:56 +0200)]
tests: gpt: exercise creating all partition types; check for msftres bug
Ensure that a newly-created partition in a GPT partition table has no
'flag' set, by default. This test also exercises setting the "name"
on a partition and creates one partition of each supported type.
* tests/t0220-gpt-msftres.sh: New file.
* tests/Makefile.am (TESTS): Add that new file.
Keshav P R [Fri, 18 Sep 2009 11:07:51 +0000 (16:37 +0530)]
gpt: don't use msftres flag for FAT/NTFS partitions
This patch corrects the bug in parted due to which any FAT(12,16,32)
or NTFS partition(s) is, by default, incorrectly set as "Microsoft
Reserved Partition" type in GPT disks. With this change, parted
defaults to setting the FAT/NTFS Partitions partition type to
Linux/Windows Basic Data Partition which will make them accessible
in Mac OS X and Windows.
Reported by hramrach@centrum.cz in <http://bugs.debian.org/441033>,
with suggested change by Osamu Aoki <osamu@debian.org>.
Keshav P R <skodabenz@rocketmail.com> wrote the above and prepared
this patch.
Jim Meyering [Thu, 24 Sep 2009 16:46:45 +0000 (18:46 +0200)]
ui: do not initialize readline (which would output!) in --script mode
Running parted with its --script (-s) option would, surprisingly,
print a few control characters if TERM were set appropriately,
and if readline and curses support were compiled in.
This fixes it not to do that.
* parted/parted.c (_init): Initialize readline support only after
parsing command line options, so we can skip it in --script mode.
* parted/ui.c (init_readline): New function.
Body extracted from ...
(init_ui): ...here.
* parted/ui.h (init_readline): Declare.
* tests/t0010-script-no-ctrl-chars.sh: New file. Test for the above.
* tests/Makefile.am (TESTS): Add that new file.
Jim Meyering [Fri, 18 Sep 2009 13:30:07 +0000 (15:30 +0200)]
loop: don't infloop
Correct a bug introduced in commit
52815877, 2007-06-15,
"Make loop_probe and loop_clobber work with sector_size > 512, too."
* libparted/labels/loop.c (loop_clobber): Initialize "i" before
the loop, not inside it.
* tests/t0400-loop-clobber-infloop.sh: Test for this.
* tests/Makefile.am (TESTS): Add it.
Jim Meyering [Thu, 17 Sep 2009 13:20:38 +0000 (15:20 +0200)]
tests: correct t1100-busy-label
This test had two problems: FS-manipulation (creates a FAT fs) even
for non-512-byte sectors, and it had a now-unexpected line in the
expected output of its final command.
* tests/t1100-busy-label.sh: Use require_512_byte_sector_size_,
so that we skip this test when in non-512-sector size mode.
Remove the now-bogus line.
Note: this test is not normally run.
In order to run it, you have to run as root and set two envvars, e.g.,
sudo make check DEVICE_TO_ERASE=/dev/sde DEVICE_TO_ERASE_SIZE=16.4MB
Jim Meyering [Sun, 13 Sep 2009 18:42:24 +0000 (20:42 +0200)]
build: update gnulib submodule to latest
Jim Meyering [Sun, 13 Sep 2009 18:44:56 +0000 (20:44 +0200)]
build: sync bootstrap from coreutils
* bootstrap: Update from coreutils.
Hans de Goede [Fri, 28 Aug 2009 15:05:55 +0000 (17:05 +0200)]
linux-commit: do not unnecessarily open partition device nodes
After patching parted with my do-not-use-BLKPG patch, I started
to get EBUSY errors on commit_to_os. Note this is not caused
by the do-not-use-BLKPG patch, this was already happening, but
parted was silently ignoring the errors (and the kernel was
not notified of the changes, which is bad). The error now
actually gets reported.
The problem turns out to be in libparted/arch/linux.c's
_flush_cache function, which walks all the partitions of the
disk and does BLKFLSBUF calls on them. This causes the following:
commit_to_os -> device_open -> fd = open /dev/sda ->
_flush_cache -> for each /dev/sda# open, ioctl, close
-> ioctl(fd, BLKRRPART) -> EBUSY
What is happening here is that the:
for each /dev/sda# open, ioctl, close
Is causing udev change events for all the /dev/sda#
nodes, which causes udev to call blkid on all these nodes
(on systems which use DeviceKit), so blkid has /dev/sda# nodes
open while BLKRRPART gets called on /dev/sda -> EBUSY.
I've checked with two independend storage subsystem kernel
developers, and /dev/sda and /dev/sda#, guarantee cache coherency
now-a-days. So there is no need to do this for 2.6, which also
eliminates the need to call _flush_cache() on device open at all.
* libparted/arch/linux.c (_have_kern26): New function.
(_flush_cache): For linux kernels 2.6 and newer, don't flush
partition devices.
(linux_open): Skip _flush_cache on newer kernels here, too.
Hans de Goede [Fri, 28 Aug 2009 08:29:00 +0000 (10:29 +0200)]
linux-commit: remove the use of the BLKPG ioctl
While testing partitionable mdraid I noticed that the kernels
view of the partition table never changes even though I was successfully
making commit_to_os() calls.
This has let to me diving into libparted's commit_to_os() code for Linux
and there are multiple issues hiding in there:
1) Parted reads /sys/block/foo/range to determine how many partitions
the device type supports and then makes BLKPG ioctl's to update the
kernels view of the partition table for partitions which fall into
this range. However for example /sys/block/sda/range contains 16,
there are 2 issue with libparted using this number:
1) scsi major's only support 15 partitions, 1 of the range of 16
is reserved for the whole device, yet libparted will try
to notify the kernel about 16 partitions if present
2) If the major's partition minor's run out, the kernel will switch
to the mdp major for the other partitions, iow range no longer
limits the number of partitions.
2) libparted assumes the user knows what he is doing, and will ignore
-ebusy errors for partitions, assuming that the user is smart enough
to only change unused partitions. Parted does this without checking
if the partitions which return ebusy actually are unchanged causing
REAL errors to get unreported (BAD, really really BAD)
3) because of 1) libparted will only sync 1 partition on /dev/md#
devices (would be 0 if not for the of by 1 bug as all md#p#
partitions use the mdp major), and it fails to even do that without
reporting an error.
###
1) we can fix by simply not checking /sys/block/foo/range, but instead
just syncing max partitions.
2) is more troublesome, we could just make -EBUSY n error,
but that may annoy / bug some users. OTOH in certain cases libparted
already falls back to BLKRRPART which will return EBUSY so users
should already be prepared to handle EBUSY
3) Could be fixed by making libparted recognize mdraid as a device type
and except mdraid from using BLKPG, like it already is doing with
DASD, but it might be better to just get rid of using BLKPG all
together. See below.
An even bigger problem IMHO is the use of the BLKPG ioctl instead of
BLKRRPART at all. What this does is tell the kernel parted's view
of the partition table and make it use that, instead of telling
the kernel to reread the partition table. According to the parted
sources this is done for the case where the kernel does not know
the disklabel type. However as soon as the system is rebooted, the
system will be using the kernel's view. So IMHO it would be much
better to always use the kernels view and just always call BLKRRPART
in commit_to_os(), this would solve all of the above issues, *and*
make the way the system views the partition table consistent between
just after running parted and after a reboot.
I've attached a patch which removes the use of the BLKPG ioctl, notice
that this also removes a lot of special case code and workarounds,
which existence to me clearly indicates that using the BLKPG ioctl is
a bad idea.
* libparted/arch/linux.c (linux_disk_commit): Remove the use of the
BLKPG ioctl.
(_blkpg_add_partition, _blkpg_part_command, _blkpg_remove_partition):
(_device_get_partition_range, _disk_sync_part_table, _have_blkpg):
(_have_devfs): Remove functions thus rendered unused.
Jim Meyering [Fri, 28 Aug 2009 16:53:39 +0000 (18:53 +0200)]
tests: make two partition-related tests work for other sector sizes
These two root-only tests would fail with the PARTED_SECTOR_SIZE envvar
set to anything other than 512. Now they also work for multiples.
* tests/t4100-dvh-partition-limits.sh: Make sector-size agnostic.
* tests/t4100-msdos-partition-limits.sh: Likewise.
Jim Meyering [Fri, 28 Aug 2009 09:47:40 +0000 (11:47 +0200)]
tests: make it easier to diagnose loop_setup_ failure
* tests/lvm-utils.sh: Don't redirect stderr to /dev/null.
Jim Meyering [Fri, 28 Aug 2009 09:30:40 +0000 (11:30 +0200)]
tests: avoid spurious failure due to extra space in diagnostic
* libparted/labels/pt-tools.c (ptt_partition_max_start_len):
Remove stray space in diagnostic that was causing the root-only
regression test, t4100-dvh-partition-limits.sh, to fail.
Jim Meyering [Fri, 28 Aug 2009 09:17:34 +0000 (11:17 +0200)]
tests: avoid spurious failure on "nodev" mounted partition
* tests/t8000-loop.sh: Skip this test if loop_setup_ fails.
* tests/t6000-dm.sh: Likewise.
Hans de Goede [Thu, 27 Aug 2009 18:16:09 +0000 (20:16 +0200)]
parted: avoid unnecessary open/close on commit, and thus udev activity
* libparted/disk.c (ped_disk_commit): Open/close the underlying file
descriptor in this function, so that callees, ped_disk_commit_to_dev
and ped_disk_commit_to_os do not each perform open/close syscalls.
This saves an open/close pair, and thus avoids unneeded udev
activity on Linux.
Before this change, when calling commit() on a ped_disk, the
following would happen:
open /dev/sda
write partition table
close /dev/sda
open /dev/sda
ioctl (BLKRRPART)
close /dev/sda
This is rather inefficient, and causes 2 udev change events to be fired
for /dev/sda (+ the change events from the BLKRRPART), causing all kind
of scanning (blkid & friends) twice.
This patch fixes things to only open the device once.
Petr Uzel [Wed, 26 Aug 2009 10:29:46 +0000 (12:29 +0200)]
dos: adjust variable declarations
* libparted/labels/dos.c (write_ext_table): Move declaration of S down
to just before its first use.
Fix alignment of lba_offset declaration.