[prev in list] [next in list] [prev in thread] [next in thread] 

List:       linux-block
Subject:    Re: [PATCH -next v2 5/9] blk-iocost: fix divide by 0 error in calc_lcoefs()
From:       Tejun Heo <tj () kernel ! org>
Date:       2022-11-30 20:20:15
Message-ID: Y4e6/wdNvZmV1c9Y () slm ! duckdns ! org
[Download RAW message or body]

On Wed, Nov 30, 2022 at 09:21:52PM +0800, Li Nan wrote:
> echo max of u64 to cost.model can cause divide by 0 error.
> 
>   # echo 8:0 rbps=18446744073709551615 > /sys/fs/cgroup/io.cost.model
> 
>   divide error: 0000 [#1] PREEMPT SMP
>   RIP: 0010:calc_lcoefs+0x4c/0xc0
>   Call Trace:
>    <TASK>
>    ioc_refresh_params+0x2b3/0x4f0
>    ioc_cost_model_write+0x3cb/0x4c0
>    ? _copy_from_iter+0x6d/0x6c0
>    ? kernfs_fop_write_iter+0xfc/0x270
>    cgroup_file_write+0xa0/0x200
>    kernfs_fop_write_iter+0x17d/0x270
>    vfs_write+0x414/0x620
>    ksys_write+0x73/0x160
>    __x64_sys_write+0x1e/0x30
>    do_syscall_64+0x35/0x80
>    entry_SYSCALL_64_after_hwframe+0x63/0xcd
> 
> calc_lcoefs() uses the input value of cost.model in DIV_ROUND_UP_ULL,
> overflow would happen if bps plus IOC_PAGE_SIZE is greater than
> ULLONG_MAX, it can cause divide by 0 error.I_LCOEF_MAX is introduced to
> prevent it.
> 
> Signed-off-by: Li Nan <linan122@huawei.com>
> ---
>  block/blk-iocost.c | 5 +++++
>  1 file changed, 5 insertions(+)
> 
> diff --git a/block/blk-iocost.c b/block/blk-iocost.c
> index 3a96cd557c47..f4a754b9d10f 100644
> --- a/block/blk-iocost.c
> +++ b/block/blk-iocost.c
> @@ -306,6 +306,9 @@ enum {
>  	IOC_PAGE_SIZE		= 1 << IOC_PAGE_SHIFT,
>  	IOC_SECT_TO_PAGE_SHIFT	= IOC_PAGE_SHIFT - SECTOR_SHIFT,
>  
> +	/* avoid overflow */
> +	I_LCOEF_MAX		= ULLONG_MAX - IOC_PAGE_SIZE,
> +
>  	/* if apart further than 16M, consider randio for linear model */
>  	LCOEF_RANDIO_PAGES	= 4096,
>  };
> @@ -3431,6 +3434,8 @@ static ssize_t ioc_cost_model_write(struct kernfs_open_file *of, char *input,
>  			goto out_unlock;
>  		}
>  
> +		if (v > I_LCOEF_MAX)
> +			goto out_unlock;

This is kinda round-about. Can you just make calc_lcoefs to not divide by
zero (e.g. just make it 1 if zero)?

Thanks.

-- 
tejun
[prev in list] [next in list] [prev in thread] [next in thread]