linux-kernel - Re: [RFC PATCH] relay: avoid relay_open_buf inproperly fails in buffer-only mode

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] [thread-next>] [day] [month] [year] [list]

Message-ID: <3b2805c0-daa1-4d2f-a3b3-4376a51ab219@linux.alibaba.com>
Date: Wed, 20 Dec 2023 16:07:24 +0800
From: Philo Lu <lulie@...ux.alibaba.com>
To: linux-kernel@...r.kernel.org
Cc: akpm@...ux-foundation.org, surenb@...gle.com, rppt@...nel.org,
 zhou.kete@....com, zhao_lei1@...erun.com, kunyu@...china.com,
 zhang.zhengming@....com, gregkh@...uxfoundation.org,
 xuanzhuo@...ux.alibaba.com, dust.li@...ux.alibaba.com,
 alibuda@...ux.alibaba.com, guwen@...ux.alibaba.com, hengqi@...ux.alibaba.com
Subject: Re: [RFC PATCH] relay: avoid relay_open_buf inproperly fails in
 buffer-only mode

Sorry, the email address of Andrew was wrongly spelled, so CC again.

On 2023/12/20 15:47, Philo Lu wrote:
> In buffer-only mode, relay_open(NULL, NULL, ...) is used to create the
> buffer first, where chan->has_base_filename is not set. Though we still
> need to call chan->cb->create_buf_file in relay_open_buf() to retrieve
> global info for global buffer, the create_buf_file callback should
> return NULL. With IS_ERR_OR_NULL() check, relay_open fails because of
> the returned NULL dentry, so this patch reverts back to the WARN_ON()
> version and add a comment for this behavior.
> 
> Here is an example after fix:
> ```
> struct dentry *my_create_buf_file(const char *filename,
>              struct dentry *parent, umode_t mode,
>              struct rchan_buf *buf, int *is_global)
> {
>      if (!filename)
>          return NULL;
> 
>      return debugfs_create_file(filename, mode, parent, buf,
>                  &relay_file_operations);
> }
> 
> relay_cb.create_buf_file = my_create_buf_file
> relay_chan = relay_open(NULL, NULL,
>                      subbuf_size, subbuf_num,
>                      &relay_cb, NULL);
> relay_late_setup_files(relay_chan, filename, parent);
> ```
> 
> But before fix, the create_buf_file callback must be something like:
> ```
> struct dentry *my_create_buf_file(const char *filename,
>              struct dentry *parent, umode_t mode,
>              struct rchan_buf *buf, int *is_global)
> {
>      if (!filename)
>          return ERR_PTR(1); // a valid ptr is necessary for relay_open
> 
>      return debugfs_create_file(filename, mode, parent, buf,
>                  &relay_file_operations);
> }
> ```
> 
> I'm not sure if this revertion proper because it may break existing use
> cases. I think we can also remove the WARN_ON check instead.
> 
> Fixes: 2c1cf00eeacb ("relay: check return of create_buf_file() properly")
> Signed-off-by: Philo Lu <lulie@...ux.alibaba.com>
> ---
>   kernel/relay.c | 3 ++-
>   1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/kernel/relay.c b/kernel/relay.c
> index 83fe0325cde1..0700745447c1 100644
> --- a/kernel/relay.c
> +++ b/kernel/relay.c
> @@ -395,7 +395,8 @@ static struct rchan_buf *relay_open_buf(struct rchan *chan, unsigned int cpu)
>   		dentry = chan->cb->create_buf_file(NULL, NULL,
>   						   S_IRUSR, buf,
>   						   &chan->is_global);
> -		if (IS_ERR_OR_NULL(dentry))
> +		/* has_base_filename not set, so dentry should be NULL */
> +		if (WARN_ON(dentry))
>   			goto free_buf;
>   	}
> 
> --
> 2.32.0.3.g01195cf9f