All of lore.kernel.org
 help / color / mirror / Atom feed
From: "xuyang2018.jy@fujitsu.com" <xuyang2018.jy@fujitsu.com>
To: Petr Vorel <pvorel@suse.cz>
Cc: "ltp@lists.linux.it" <ltp@lists.linux.it>,
	"martin.doucha@suse.com" <martin.doucha@suse.com>
Subject: Re: [LTP] quotactl0{2, 3, 5, 7}.c dependency problems on older kernels
Date: Fri, 18 Mar 2022 09:26:12 +0000	[thread overview]
Message-ID: <62345060.1070604@fujitsu.com> (raw)
In-Reply-To: <YjRH9NOKME95aABF@pevik>

Hi Petr
> Hi Xu,
>
>> Hi Petr
>>> Hi Xu,
>
>>> actually quotactl0{2,3,5,7}.c and quotactl03.c fail on only on v4.5:
>
>>> tst_kconfig.c:82: TINFO: Parsing kernel config '/.config'
>>> tst_device.c:88: TINFO: Found free device 0 '/dev/loop0'
>>> tst_test.c:996: TINFO: Formatting /dev/loop0 with xfs opts='' extra opts=''
>>> [  271.093416] XFS (loop0): Superblock has unknown read-only compatible features (0x4) enabled.
>>> [  271.096902] XFS (loop0): Attempted to mount read-only compatible filesystem read-write.
>>> [  271.100128] XFS (loop0): Filesystem can only be safely mounted read only.
>>> [  271.101852] XFS (loop0): SB validate failed with error -22.
>>> tst_test.c:1010: TBROK: mount(/dev/loop0, mntpoint, xfs, 0, 0x41d6a0) failed: EINVAL (22)
>> That is because you use a newer xfsprogs (it provides newer mkfs.xfs and
>> enable reflink feature by default[1]). But older kernel can't regcozine
>> this feature bits[2](it supports reflink feature bits since kernel
>> 4.9-rc1), then mount failed.
>
>> [1]https://git.kernel.org/pub/scm/fs/xfs/xfsprogs-dev.git/commit/?id=ec1b42e6d58dcb7cfc5de6fc825832944997e176
>> [2]https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit?id=f0ec1b8ef11df0a51954df7e3ff3ca4aadb0d34b
>
>> I don't think this is a problem, user should know its kernel version
>> and xfsprogs vesion. Linux distribution has sloved this well except you
>> used yourself compiled kernel.
>
> TL;DR: Understand and we can stop debugging to not waste time.
>
> Giving userspace tool needs to match kernel brings extra barrier when bisecting
> in the history :). I suppose there is no list to match these (actually it'd be
> better to ask on linux-xfs). I'd have to experiment with slightly older ones -
> I'm using mkfs.xfs 5.14.2 from openSUSE Tumbleweed (i.e. latest stable).
>
>>> Tested on config:
>>> grep -e QUOTA -e QFMT /.config
>>> CONFIG_XFS_QUOTA=y
>>> CONFIG_QUOTA=y
>>> CONFIG_QUOTA_NETLINK_INTERFACE=y
>>> # CONFIG_PRINT_QUOTA_WARNING is not set
>>> # CONFIG_QUOTA_DEBUG is not set
>>> CONFIG_QUOTA_TREE=y
>>> # CONFIG_QFMT_V1 is not set
>>> CONFIG_QFMT_V2=y
>>> CONFIG_QUOTACTL=y
>>> CONFIG_QUOTACTL_COMPAT=y
>
>>> But not work either on v4.10:
>>> [  299.235456] XFS (loop0): EXPERIMENTAL reflink feature enabled. Use at your own risk!
>>> [  299.239463] XFS (loop0): Mounting V5 Filesystem
>>> [  299.240950] XFS (loop0): Log size 1368 blocks too small, minimum size is 1580 blocks
>> In xfstests, we use -lsize parameter[3] or use a value then get the
>> correct value from error log[4]. reflink feature will increase log
>> size[5], so we can disable reflink feature by using mkfs.xfs -m
>> reflink=0 /dev/loop0 (it also need mkfs.xfs support, so we may need a
>> mkfs.xfs cmd version check[5], if meet this version, use mkfs.xfs -m
>> reflink=0, if less than this version, we can ignore this version and
>> don't use mkfs.xfs extra option).
>
>> Firstly, can you try mkfs.xfs -m reflink=0 way whether fix these  two
>> problems?
>
> Yes, adding .dev_fs_opts:
> @@ -84,6 +88,12 @@ static struct tst_test test = {
>          .test_all = verify_quota,
>          .format_device = 1,
>          .dev_fs_type = "xfs",
> +       .dev_fs_opts = (const char *const[]){
> +               "-m",
> +               "reflink=0",
> +               NULL,
> +       },
> +
>
> ... helps to quotactl0{2,3,5,7}.c helps to all:
>
> quotactl0{2,3,5}.c TPASS/TCONF:
> tst_test.c:996: TINFO: Formatting /dev/loop0 with xfs opts='-m reflink=0' extra opts=''
> [   55.193262] XFS (loop0): Mounting V5 Filesystem
> [   55.197574] XFS (loop0): Ending clean mount
> [   55.199560] XFS (loop0): Quotacheck needed: Please wait.
> [   55.202735] XFS (loop0): Quotacheck: Done.
> tst_test.c:1456: TINFO: Timeout per run is 0h 05m 00s
> quotactl_syscall_var.h:27: TINFO: Test quotactl()
> quotactl02.c:119: TINFO: Test #0: QCMD(Q_XGETQSTAT, USRQUOTA) off
> quotactl02.h:64: TPASS: quotactl() succeeded to turn off xfs quota and get xfs quota off status for user
> ...
> ../../../../include/lapi/quotactl.h:18: TCONF: syscall(443) __NR_quotactl_fd not supported on your arch
>
> quotactl07.c fails on Q_XQUOTARM, but that's correct (missing 3dd4d40b4208 from v5.5)
>
> tst_test.c:996: TINFO: Formatting /dev/loop0 with xfs opts='-m reflink=0' extra opts=''
> ...
> quotactl07.c:37: TPASS: do_quotactl(Q_XQUOTARM,valid_type) passed
> quotactl07.c:41: TPASS: Q_XQUOTARM to free space, delta(1)
> quotactl07.c:45: TFAIL: do_quotactl(Q_XQUOTARM, invalid_type) succeeded
> [  116.407282] XFS (loop0): Unmounting Filesystem
> tst_test.c:1456: TINFO: Timeout per run is 0h 05m 00s
> ...
> ../../../../include/lapi/quotactl.h:18: TCONF: syscall(443) __NR_quotactl_fd not supported on your arch
>
> =>  not sure if we want to add this "-m reflink=0" to these tests (into
> quotactl_syscall_var.h, which they all use).
xfsprogs supports -m reflink after 2016.10, some old distributions 
xfsprogs doesn't support it ie centos7.9ga.

So if we add this, it will TCONF on these old linux distributions.

I think we need a mkfs.xfs wrapper in quotactl_syscall_var.h, this 
wrapper does the following thing:
1. mkfs.xfs -m reflink=0 succeed then return
2. mkfs.xfs -m reflink=0 failed then use mkfs.xfs to format, then return


ps:Another way is using MKFS_OPTIONS environment like xfstests, then we 
can refer to filesytem type to get value ie xfs MKFS_OPTIONS gets from 
XFS_MKFS_OPTIONS[1]. so the style of a complete mkfs option in case is 
that XFS_MKFS_OPTIONS + dev_fs_opts. But the former way is simple and 
enough to fix.

[1]https://git.kernel.org/pub/scm/fs/xfs/xfstests-dev.git/tree/common/config#n420

Best Regards
Yang Xu
>
> Kind regards,
> Petr
>
>> [3]https://git.kernel.org/pub/scm/fs/xfs/xfstests-dev.git/commit/?id=7e07c4bebb08414ca8228ea91bdf57828fd4d110
>> [4]https://git.kernel.org/pub/scm/fs/xfs/xfstests-dev.git/tree/common/xfs#n81
>> [5]https://git.kernel.org/pub/scm/fs/xfs/xfsprogs-dev.git/commit/?id=a5132d9b3634fb6436d1f06642ceda82e64ea2f5
>
>> Best Regards
>> Yang Xu

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

  reply	other threads:[~2022-03-18  9:26 UTC|newest]

Thread overview: 7+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-03-17 18:31 [LTP] quotactl0{2,3,5,7}.c dependency problems on older kernels Petr Vorel
2022-03-18  3:16 ` [LTP] quotactl0{2, 3, 5, 7}.c " xuyang2018.jy
2022-03-18  3:44   ` xuyang2018.jy
2022-03-18 12:23     ` Petr Vorel
2022-03-18  8:51   ` Petr Vorel
2022-03-18  9:26     ` xuyang2018.jy [this message]
2022-03-18  9:59       ` Petr Vorel

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=62345060.1070604@fujitsu.com \
    --to=xuyang2018.jy@fujitsu.com \
    --cc=ltp@lists.linux.it \
    --cc=martin.doucha@suse.com \
    --cc=pvorel@suse.cz \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.