June 2022 - Kernel - mailweb.openeuler.org

[PATCH openEuler-1.0-LTS] xfs: show the proper user quota options
by tangbin 02 Jun '22

02 Jun '22

From: Kaixu Xia <kaixuxia(a)tencent.com> mainline inclusion from mainline-v5.11-rc6 commit 237d7887ae723af7d978e8b9a385fdff416f357b category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/I5AG18 CVE: NA -------------------------------- The quota option 'usrquota' should be shown if both the XFS_UQUOTA_ACCT and XFS_UQUOTA_ENFD flags are set. The option 'uqnoenforce' should be shown when only the XFS_UQUOTA_ACCT flag is set. The current code logic seems wrong, Fix it and show proper options. Signed-off-by: Kaixu Xia <kaixuxia(a)tencent.com> Reviewed-by: Darrick J. Wong <darrick.wong(a)oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong(a)oracle.com> Signed-off-by: tangbin <tangbin(a)cmss.chinamobile.com> --- fs/xfs/xfs_super.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c index e126bc6cdeb8..5d8e3a4d2671 100644 --- a/fs/xfs/xfs_super.c +++ b/fs/xfs/xfs_super.c @@ -505,10 +505,12 @@ xfs_showargs( seq_printf(m, ",swidth=%d", (int)XFS_FSB_TO_BB(mp, mp->m_swidth)); - if (mp->m_qflags & (XFS_UQUOTA_ACCT|XFS_UQUOTA_ENFD)) - seq_puts(m, ",usrquota"); - else if (mp->m_qflags & XFS_UQUOTA_ACCT) - seq_puts(m, ",uqnoenforce"); + if (mp->m_qflags & XFS_UQUOTA_ACCT) { + if (mp->m_qflags & XFS_UQUOTA_ENFD) + seq_puts(m, ",usrquota"); + else + seq_puts(m, ",uqnoenforce"); + } if (mp->m_qflags & XFS_PQUOTA_ACCT) { if (mp->m_qflags & XFS_PQUOTA_ENFD) -- 2.18.4

1 0

[PATCH] xfs: show the proper user quota options
by tangbin 02 Jun '22

02 Jun '22

From: Kaixu Xia <kaixuxia(a)tencent.com> mainline inclusion from mainline-v5.11-rc6 commit 237d7887ae723af7d978e8b9a385fdff416f357b category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/I5AG18 CVE: NA -------------------------------- The quota option 'usrquota' should be shown if both the XFS_UQUOTA_ACCT and XFS_UQUOTA_ENFD flags are set. The option 'uqnoenforce' should be shown when only the XFS_UQUOTA_ACCT flag is set. The current code logic seems wrong, Fix it and show proper options. Signed-off-by: Kaixu Xia <kaixuxia(a)tencent.com> Reviewed-by: Darrick J. Wong <darrick.wong(a)oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong(a)oracle.com> Signed-off-by: tangbin <tangbin(a)cmss.chinamobile.com> --- fs/xfs/xfs_super.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c index e126bc6cdeb8..5d8e3a4d2671 100644 --- a/fs/xfs/xfs_super.c +++ b/fs/xfs/xfs_super.c @@ -505,10 +505,12 @@ xfs_showargs( seq_printf(m, ",swidth=%d", (int)XFS_FSB_TO_BB(mp, mp->m_swidth)); - if (mp->m_qflags & (XFS_UQUOTA_ACCT|XFS_UQUOTA_ENFD)) - seq_puts(m, ",usrquota"); - else if (mp->m_qflags & XFS_UQUOTA_ACCT) - seq_puts(m, ",uqnoenforce"); + if (mp->m_qflags & XFS_UQUOTA_ACCT) { + if (mp->m_qflags & XFS_UQUOTA_ENFD) + seq_puts(m, ",usrquota"); + else + seq_puts(m, ",uqnoenforce"); + } if (mp->m_qflags & XFS_PQUOTA_ACCT) { if (mp->m_qflags & XFS_PQUOTA_ENFD) -- 2.18.4

1 0

Re: [PATCH OLK-5.10 1/4] eulerfs: add socket_num to eufs_sb_info
by Gou Hao 02 Jun '22

02 Jun '22

在 2022/6/2 17:10, Yu Kuai 写道: > 在 2022/06/02 10:50, Gou Hao 写道: >> uniontech inclusion >> category: bugfix >> bugzilla: NA >> CVE: NA >> >> ---------------------- >> >> We should not get the socket num in init_eufs_fs, becuse >> system usually supports cpu hotplug,so we should get socket >> num in dep_int. > Hi, > > What if cpu is offline during mount, and online afterwards? > Well, in this case, we have to add a monitoring mechanism to dynamically adjust the number of threads according to the online status of the CPU. Or there are other better ways ... > Thanks, > Kuai >> >> Socket num is not necessarily the same when file system >> mount,so it shoule be eufs_sb_info's member. >> >> Signed-off-by: Gou Hao <gouhao(a)uniontech.com> >> --- >> fs/eulerfs/dep.c | 28 ++++++++++++++++++++++------ >> fs/eulerfs/euler.h | 2 -- >> fs/eulerfs/euler_def.h | 1 + >> fs/eulerfs/super.c | 15 +-------------- >> 4 files changed, 24 insertions(+), 22 deletions(-) > -- thanks, Gou Hao

1 0

Re: [PATCH OLK-5.10 4/4] eulerfs: fix potential sbi->persisters free error
by Gou Hao 02 Jun '22

02 Jun '22

在 2022/6/2 17:15, Yu Kuai 写道: > 在 2022/06/02 10:51, Gou Hao 写道: >> uniontech inclusion >> category: bugfix >> bugzilla: NA >> CVE: NA >> >> ------------------- >> >> After alloc the sbi->persisters memory, dep_init >> will call dep_fini when error happened.Because >> sbi->persisters is not set to 0, kthread_stop >> may happened error. > > Looks good to me, just one language fixup: > > sbi->persisters is not set to 0 > -> dep_fini() can be called with sbi->persisters[] uninitialized, thus > kthread_stop() can be called with random value. > ok, thank you : ) > Thanks, > Kuai >> >> Signed-off-by: Gou Hao <gouhao(a)uniontech.com> >> --- >> fs/eulerfs/dep.c | 4 ++-- >> 1 file changed, 2 insertions(+), 2 deletions(-) >> >> diff --git a/fs/eulerfs/dep.c b/fs/eulerfs/dep.c >> index da826a18fa31..62fd1376f980 100644 >> --- a/fs/eulerfs/dep.c >> +++ b/fs/eulerfs/dep.c >> @@ -728,8 +728,8 @@ int dep_init(struct super_block *sb) >> node_num++; >> sbi->persister_num = node_num * persisters_per_socket; >> - >> - sbi->persisters = kmalloc(sizeof(struct task_struct *) * >> + >> + sbi->persisters = kzalloc(sizeof(struct task_struct *) * >> sbi->persister_num, GFP_KERNEL); >> if (!sbi->persisters) { >> err = -ENOMEM; >> > -- thanks, Gou Hao

1 0

[PATCH OLK-5.10 0/4] eulerfs: improve the creation of persister
by Gou Hao 02 Jun '22

02 Jun '22

Patch series "improve the creation of persister". The main modification is to create a persistent thread only when a NUMA node has an online CPU at least, instead of on all NUMA nodes. In the current code, the number of NUMA nodes is obtained when the module is loaded, and the same number of persistent threads are created when the file system is mounted. However, we should ignore NUMA nodes that do not have an online CPU. So we should decide how many threads to create according to the online CPU condition when initializing persist. Patch4 is a bugfix about dep_init(). Gou Hao (4): eulerfs: add socket_num to eufs_sb_info eulerfs: create persister on node when it has online cpu eulerfs: record persister num eulerfs: fix potential sbi->persisters free error fs/eulerfs/dep.c | 48 +++++++++++++++++++++++++++--------------- fs/eulerfs/euler.h | 2 -- fs/eulerfs/euler_def.h | 1 + fs/eulerfs/super.c | 21 ++++-------------- 4 files changed, 36 insertions(+), 36 deletions(-) -- 2.20.1

1 4