linux-kernel - Re: [PATCH 1/2] hp-wmi: fix unregister order in hp_wmi_rfkill

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: <56DEF6B0.9050902@maciej.szmigiero.name>
Date:	Tue, 08 Mar 2016 16:58:40 +0100
From:	"Maciej S. Szmigiero" <mail@...iej.szmigiero.name>
To:	Darren Hart <dvhart@...radead.org>
CC:	platform-driver-x86@...r.kernel.org,
	linux-kernel <linux-kernel@...r.kernel.org>,
	Kirill Tkhai <tkhai@...dex.ru>
Subject: Re: [PATCH 1/2] hp-wmi: fix unregister order in hp_wmi_rfkill_setup()
 once again

Hi Darren,

Thanks for review, see also my comments below.

On 08.03.2016 13:39, Darren Hart wrote:
> On Sun, Mar 06, 2016 at 11:38:36PM +0100, Maciej S. Szmigiero wrote:
>> --- a/drivers/platform/x86/hp-wmi.c
>> +++ b/drivers/platform/x86/hp-wmi.c
>> @@ -746,7 +746,7 @@ static int __init hp_wmi_rfkill_setup(struct platform_device *device)
>>  						(void *) HPWMI_BLUETOOTH);
>>  		if (!bluetooth_rfkill) {
>>  			err = -ENOMEM;
>> -			goto register_wifi_error;
>> +			goto register_bluetooth_error;
> 
> In this and all cases below, the goto label should match the situation, jumping
> to register_bluetooth_error would be incorrect as we experienced a wifi error. 

Here we experienced an BT error - BT rkill allocation failed,
so jump is to "register_bluetooth_error".

The second jump to "register_bluetooth_error" is in another case of BT error:
when its rfkill registration failed.

It is the same label since if BT rfkill allocation had failed
rfkill_destroy(bluetooth_rfkill) call in cleanup does nothing but we still
need to possibly unregister WiFi rfkill that might have been registered in
a previews block (and then fall through to next label to destroy WiFi rfkill).

It would be possible to have separate jump labels skipping unnecessary
rfkill_destroy() calls on allocation failure, but this would mean
that we would have 7 labels in such small block of cleanup code.

> A better solution would be to reorder the labels in the exit block
> such that they enforce the necessary reverse order.

Cleanup labels already are in reverse order with regard to registration:
Registration:
1) WiFi,
2) BT,
3) WWAN,
5) GPS.

Cleanup:
1) GPS,
2) WWAN,
3) BT,
4) WiFi.

Best regards,
Maciej Szmigiero