ploop.spec: enhance BuildRoot as per EPEL guidelineshttp://fedoraproject.org/wiki/EPEL/GuidelinesAndPolicies#BuildRoot_tag
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
ploop.spec: force vzctl >= 4.1.2vzctl versions older than 4.1.2 has a buffer overflow that strikes
when using ploop-lib 1.6. Since ploop-lib should not require vzctl,
use Conflicts tag to say what we want to say.
See more details at
http://git.openvz.org/?p=vzctl;a=commit;h=658010cf
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Revert "ploop.spec: require libs of the proper arch"Apparently in RHEL5 (which we use to build this package)
this macro (%_isa) is not supported, so we end up with a
broken dependency.
This reverts commit 3f07010bd930837082de32ba012411b28a6440be.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
[build] fix check-api logicApparently we need to sort both lists of functions before trying to
run diff on those, otherwise diff complains about function positions,
which is not what we want.
So change the logic of
1. gensym.sh to emit just sorted list of functions
2. include/Makefile's check-api target to generate cur-dynload.h
as a sorted list of functions.
diff will now ignore the differences in positions.
NEVER...
ploop.spec: require our static e2fsresizeThis is an addition to commit 2a50fb2. See the description
in there.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
dynload.h: add more external functions* create_snapshot_ex() introduced by commit b2cf1f4
* complete_running_operation() made public in commit 9933856
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
[build] introduce hook to check if dynload.h is up to dateWith this patch, make will check if dynload.h is up to date,
and exit with an error if it's not. In such case, you need
to manually add missing functions to dynload.h, near the end
of struct ploop_functions.
The case when we remove an external function is not yet
supported.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
include/dynload.h: addThis file was autogenerated before. We can't do that now when we
are adding more functions to libploop.h, since ordering is important.
So, this commit adds a static version of dynload.h as it appeared
in ploop-1.5 (with slight changes such as added/removed comments
and preprocessor inclusion guard). The next commit will introduce
a way to check if the file is up-to-date.
Signed-off-by: Kir Kol...
Call custom /usr/libexec/resize2fs if availableresize2fs as shipped in RHEL6 distro does not have support for
EXT4_IOC_RESIZE_FS ioctl which greatly improves resizing speed.
Such support appeared in e2fsprogs-1.42, released Nov 29, 2011.
(Note that standard RHEL6 kernel doesn't have this support either,
it comes with OpenVZ RHEL6-based kernel).
Reference: PSBM-14853
Reference: ploop@829441
alloc_diskdescriptor(): log error if calloc() failsOtherwise functions such as ploop_alloc_disk_descr() may
fail silently, and then ploop_get_last_error() will show
something irrelevant.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
lib/ploop.h: remove PL_EXT from couple of protosThese two functions are not used by tools and were mistakenly
marked with PL_EXT in commit 85fc03a. Fix it.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Revert "add_delta(): prevent multiple mounts"As Maxim Patlasov puts it:
> We regard "ploop cmd ..." as low-level tool for developers,
> testers and admins. Guy should know what he's doing.
>
> Kir Kolyshkin wrote:
>
>> Still, there is a possibility of race here, and we can't use
>> ploop_global_lock() because it is already used in
>> ploop_find_dev(), so suggestions are welcome.
>
> This the main reason to NACK the patch. It'd give the f...
__ploop_discard(): fix exit code after 809436Before commit ploop@809436 the code was:
ret = ploop_trim(mount_point, minlen_b, cluster);
...
exit(ret < 0);
and since ploop_trim() only returns 0 or -1 it was all correct.
After commit ploop@809436 ret can get value from blk_discard()
which can return SYSEXIT_DEVIOC so the (ret < 0) check is no
longer valid.
Fix by using (ret != 0)
Reference: ploop@821763
Signed-off-by: Kir Kolyshkin ...
[style] resize_gpt_partition(): do not init retNo need to initialize ret as we do that at the very beginning
of the function.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
ploop_resize_image(): fix compiler warning ploop.c: In function 'ploop_resize_image':
ploop.c:1626:6: error: 'new_fs_size' may be used uninitialized in this function [-Werror=maybe-uninitialized]
ploop.c:1663:8: note: 'new_fs_size' was declared here
Introduced by commit b8fbdad
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
ploop_resize_image(): do not return -1Fix a few cases in ploop_resize_image() in which we return -1.
We should not do this as the value is used as an exit code
from ploop resize command.
This, together with the 10 or so previous patches, should significantly
decrease the probability of doing exit(-1) from ploop resize. I am not
sure though I eliminated all such cases.
Reference: ploop@821373
Signed-off-by: Kir Kolyshkin <kir@ope...
resize_gpt_partition(): do not return -1, more exit codes1. We should not return -1 from resize_gpt_partition() since the
value can eventually be used as exit code from ploop resize
and ploop balloon discard.
2. If we can return a value less generic than SYSEXIT_CHANGE_GPT,
do that
Reference: ploop@821372
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
resize_fs(): do not return -1resize_fs() return value can be eventually used as ploop resize
exit code so we end up doing exit(-1).
Fix by returning SYSEXIT_SYS.
Reference: ploop@821371
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
make_fs(): do not return -1This value is eventually used as ploop init exit code
so we can't return -1 from here. Use SYSEXIT_MKFS.
Reference: ploop@821370
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
dumpe2fs(): do not return -1Return code from dumpe2fs() can be returned from ploop_resize_image()
and is then used as ploop resize exit code, leading to exit(-1).
Fix by returning SYSEXIT_SYS from dumpe2fs() in case of errors.
Also, now we can also return this value directly from shrink_device(),
which return value can also be returned from ploop_resize_image().
So, this patch fixes two cases where 'ploop resize' can d...
ploop balloon *: fix exit codeBy not returning -1 from get_balloon(), and returning SYSEXIT_PARAM
instead, we fix about 8 "exit(-1)" cases with ploop balloon commands.
Reference: ploop@821366
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
ploop ballon: fixed a number of exit(-1) casesWe should not return -1 from main(), exit codes are positive
numbers.
Fix by using SYSEXIT_PARAM in most cases, and SYSEXIT_SYS for the case
when ploop_lib_init() fails.
Reference: ploop@821365
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Introduce and use SYSEXIT_FSCKIn case fsck on an underlying file system failed,
return the appropriate error code.
In particular, this fixes two cases when ploop_resize_image()
was returning -1 which is wrong because its return value is
used as an exit code from "ploop resize".
Reference: ploop@821364
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
ploop_resize_image(): fix an error codeWe should not return -1 from this function because
its return code is used as exit code for 'ploop resize'.
In case we don't like DiskDescriptor.xml, the proper code
to return is SYSEXIT_DISKDESCR.
Reference: ploop@821363
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
ploop.spec: require util-linuxWe use /sbin/mkfs which is in util-linux, so let's require it.
Reference: ploop@821189
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
ploop.spec: require e2fsprogsWe use resize2fs binary, plus since commit 7e47931 we use tune2fs,
plus since commit 0badb9d we use e2fsck and dumpe2fs from the library.
These binaries come from e2fsprogs package so let's require it
from the spec file.
Reference: ploop@821188
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Fix ploop_get_info() for a particular caseUpdate .statfs info in case file does not exist
and ploop device is stopped.
Reference: PSBM-16977
Reference: ploop@821111
Fixed comments in ploop_resize_image()After commit 2bb0fb3:
1 Use C-style comments for long comments
2 childs -> children
3 alligment -> alignment
After commit 83c51b1:
4 it's 4x not 2x now, fix comment accordingly
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Fix fail to shrinkAlign size supplied to resize2fs to 4k
Increase delta to (4 * blocksize) as a workaround fo future investigation
Reference: PSBM-15919
Reference: ploop@811218
Kir Kolyshkin