Re: [PATCH] arm64: fix oops in concurrently setting insn_emulation sysctls

[Will Deacon <will@kernel.org>]

On Fri, Jul 01, 2022 at 02:27:17AM +0000, haibinzhang(张海斌) wrote:
> How to reproduce:
>     launch two shell executions:
>        #!/bin/bash
>        while [ 1 ];
>        do
>            echo 1 > /proc/sys/abi/swp
>        done
> 
> Oops info:
>     Unable to handle kernel NULL pointer dereference at virtual address 0000000000000010
>     Internal error: Oops: 96000006 [#1] SMP
>     Call trace:
>     update_insn_emulation_mode+0xc0/0x148
>     emulation_proc_handler+0x64/0xb8
>     proc_sys_call_handler+0x9c/0xf8
>     proc_sys_write+0x18/0x20
>     __vfs_write+0x20/0x48
>     vfs_write+0xe4/0x1d0
>     ksys_write+0x70/0xf8
>     __arm64_sys_write+0x20/0x28
>     el0_svc_common.constprop.0+0x7c/0x1c0
>     el0_svc_handler+0x2c/0xa0
>     el0_svc+0x8/0x200
> 
> emulation_proc_handler changes table->data for proc_dointvec_minmax
> and so it isn't allowed to reenter before restoring table->data,
> which isn't right now.
> To fix this issue, Add mutal exclusion covering related code section.

typo: mutual

> Signed-off-by: Haibin Zhang <haibinzhang@tencent.com>
> ---
>  arch/arm64/kernel/armv8_deprecated.c | 9 +++++++--
>  1 file changed, 7 insertions(+), 2 deletions(-)
> 
> diff --git a/arch/arm64/kernel/armv8_deprecated.c b/arch/arm64/kernel/armv8_deprecated.c
> index 6875a16..c519792 100644
> --- a/arch/arm64/kernel/armv8_deprecated.c
> +++ b/arch/arm64/kernel/armv8_deprecated.c
> @@ -207,8 +207,12 @@ static int emulation_proc_handler(struct ctl_table *table, int write,
>  				  loff_t *ppos)
>  {
>  	int ret = 0;
> -	struct insn_emulation *insn = (struct insn_emulation *) table->data;
> -	enum insn_emulation_mode prev_mode = insn->current_mode;
> +	struct insn_emulation *insn;
> +	enum insn_emulation_mode prev_mode;
> +
> +	raw_spin_lock(&insn_emulation_lock);
> +	insn = (struct insn_emulation *) table->data;
> +	prev_mode = insn->current_mode;
>  
>  	table->data = &insn->current_mode;
>  	ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
> @@ -224,6 +228,7 @@ static int emulation_proc_handler(struct ctl_table *table, int write,
>  	}
>  ret:
>  	table->data = insn;
> +	raw_spin_unlock(&insn_emulation_lock);
>  	return ret;

This looks very similar to the patch previously posted here:

https://lore.kernel.org/r/20220128090324.2727688-1-hewenliang4@huawei.com

but Catalin's suggestion was ignored:

https://lore.kernel.org/all/Yf0dxon1d07rzxZH@arm.com/

Please can have you send a v2 along the line that he suggested?

I also think a mutex is probably better than a spinlock given that we
can end up cross-calling in the proc handler.

Thanks,

Will