lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <d7b3ad0c-6474-459d-855d-c1c435a2bb17@kernel.org>
Date: Wed, 11 Sep 2024 15:12:25 +0800
From: Chao Yu <chao@...nel.org>
To: wangzijie1@...or.com
Cc: chao@...nel.org, jaegeuk@...nel.org, linux-kernel@...r.kernel.org,
 linux-f2fs-devel@...ts.sourceforge.net
Subject: Re: [f2fs-dev] [RFC PATCH] f2fs: don't set SBI_QUOTA_NEED_REPAIR flag
 if receive SIGKILL

On 2024/9/4 11:49, Chao Yu via Linux-f2fs-devel wrote:
> On 2024/9/3 14:06, wangzijie wrote:
>> From: Chao Yu via Linux-f2fs-devel <linux-f2fs-devel@...ts.sourceforge.net>
>>
>>> On 2024/8/27 14:22, wangzijie wrote:
>>>> Thread A
>>>> -dquot_initialize
>>>>   -dqget
>>>>    -f2fs_dquot_acquire
>>>>     -v2_read_dquot
>>>>      -qtree_read_dquot
>>>>       -find_tree_dqentry
>>>>        -f2fs_quota_read
>>>>         -read_cache_page_gfp
>>>>          -do_read_cache_folio
>>>>           -fiemap_read_folio
>>>>            -folio_wait_locked_killable
>>>>             -receive SIGKILL : return -EINTR
>>>>         -set SBI_QUOTA_NEED_REPAIR
>>>>     -set SBI_QUOTA_NEED_REPAIR
>>>>
>>>> When calling read_cache_page_gfp in quota read, thread may receive SIGKILL and
>>>> set SBI_QUOTA_NEED_REPAIR, should we set SBI_QUOTA_NEED_REPAIR in this error path?
>>>
>>> f2fs_quota_read() can be called in a lot of contexts, can we just ignore -EINTR
>>> for f2fs_dquot_initialize() case?
>>>
>>> Thanks,
>>
>> Yes, in many contexts f2fs_quota_read() can be called and may return -EINTR, we need to ignore this errno for more cases. If we need to do so, I will check it and resend patch.
>> Or do you have other suggestions to avoid unnecessary SBI_QUOTA_NEED_REPAIR flag set?
> 
> How about this?
> 
> ---
>   fs/f2fs/f2fs.h  |  1 +
>   fs/f2fs/inode.c |  3 +--
>   fs/f2fs/super.c | 17 +++++++++++++----
>   3 files changed, 15 insertions(+), 6 deletions(-)
> 
> diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
> index dfed1974eda5..a1704a19dfe9 100644
> --- a/fs/f2fs/f2fs.h
> +++ b/fs/f2fs/f2fs.h
> @@ -810,6 +810,7 @@ enum {
>       FI_ATOMIC_DIRTIED,    /* indicate atomic file is dirtied */
>       FI_ATOMIC_REPLACE,    /* indicate atomic replace */
>       FI_OPENED_FILE,        /* indicate file has been opened */
> +    FI_INIT_DQUOT,        /* indicate it's initializing dquot */
>       FI_MAX,            /* max flag, never be used */
>   };
> 
> diff --git a/fs/f2fs/inode.c b/fs/f2fs/inode.c
> index 008f01348afa..b1dbaeda306f 100644
> --- a/fs/f2fs/inode.c
> +++ b/fs/f2fs/inode.c
> @@ -827,8 +827,7 @@ void f2fs_evict_inode(struct inode *inode)
> 
>       err = f2fs_dquot_initialize(inode);
>       if (err) {
> -        if (err != -EINTR)
> -            set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
> +        set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
>           err = 0;
>       }
> 
> diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
> index 8e29aba4b7a4..e774bdf875b2 100644
> --- a/fs/f2fs/super.c
> +++ b/fs/f2fs/super.c
> @@ -2644,8 +2644,11 @@ static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
>               if (PTR_ERR(page) == -ENOMEM) {
>                   memalloc_retry_wait(GFP_NOFS);
>                   goto repeat;
> -            } else if (PTR_ERR(page) != -EINTR)
> -                set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
> +            } else if (PTR_ERR(page) == -EINTR &&
> +                is_inode_flag_set(inode, FI_INIT_DQUOT)) {

Current inode is quota inode, it's not the same inode we tagged
w/ FI_INIT_DQUOT, so please ignore this diff...

Thanks,

> +                return PTR_ERR(page);
> +            }
> +            set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
>               return PTR_ERR(page);
>           }
> 
> @@ -2721,10 +2724,16 @@ static ssize_t f2fs_quota_write(struct super_block *sb, int type,
> 
>   int f2fs_dquot_initialize(struct inode *inode)
>   {
> +    int ret;
> +
>       if (time_to_inject(F2FS_I_SB(inode), FAULT_DQUOT_INIT))
>           return -ESRCH;
> 
> -    return dquot_initialize(inode);
> +    set_inode_flag(inode, FI_INIT_DQUOT);
> +    ret = dquot_initialize(inode);
> +    clear_inode_flag(inode, FI_INIT_DQUOT);
> +
> +    return ret;
>   }
> 
>   static struct dquot __rcu **f2fs_get_dquots(struct inode *inode)
> @@ -3064,7 +3073,7 @@ static int f2fs_dquot_acquire(struct dquot *dquot)
> 
>       f2fs_down_read(&sbi->quota_sem);
>       ret = dquot_acquire(dquot);
> -    if (ret < 0 && ret != -EINTR)
> +    if (ret < 0)
>           set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
>       f2fs_up_read(&sbi->quota_sem);
>       return ret;


Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ