netdev - Re: [PATCH v2] ravb: Fixed the problem that rmmod can not be done

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: <97569264-fcf8-58cb-3ce7-9d569ad176e5@gmail.com>
Date:   Fri, 31 Jul 2020 20:45:02 +0300
From:   Sergei Shtylyov <sergei.shtylyov@...il.com>
To:     Yoshihiro Shimoda <yoshihiro.shimoda.uh@...esas.com>,
        Yuusuke Ashizuka <ashiduka@...itsu.com>
Cc:     "netdev@...r.kernel.org" <netdev@...r.kernel.org>,
        "linux-renesas-soc@...r.kernel.org" 
        <linux-renesas-soc@...r.kernel.org>
Subject: Re: [PATCH v2] ravb: Fixed the problem that rmmod can not be done

Hello!


On 7/31/20 9:43 AM, Yoshihiro Shimoda wrote:

>>>> From: Yuusuke Ashizuka, Sent: Thursday, July 30, 2020 7:02 PM
>>>> Subject: [PATCH v2] ravb: Fixed the problem that rmmod can not be done
>>>
>>> Thank you for the patch! I found a similar patch for another driver [1].
>>
>>    It's not the same case -- that driver hadn't had the MDIO release code at all
>> before that patch.
> 
> You're correct. I didn't realized it...

   The patch description was somewhat incomplete there...

>>> So, we should apply this patch to the ravb driver.
>>
>>    I believe the driver is innocent. :-)
> 
> I hope so :)

   Looks like I was wrong in this case. It's very fortunate that the MDIO bitbang
is not as popular as I thought.

> <snip>
>>>> $ lsmod
>>>> Module                  Size  Used by
>>>> ravb                   40960  1
>>>> $ rmmod ravb
>>>> rmmod: ERROR: Module ravb is in use
>>
>>    Shouldn't the driver core call the remove() method for the affected devices
>> first, before checking the refcount?
> 
> In this case, an mii bus of "mdiobb_ops bb_ops" is affected "device" by the ravb driver.
> And the ravb driver sets the owner of mii bus as THIS_MODULE like below:
> 
> static struct mdiobb_ops bb_ops = {
>         .owner = THIS_MODULE,
>         .set_mdc = ravb_set_mdc,
>         .set_mdio_dir = ravb_set_mdio_dir,
>         .set_mdio_data = ravb_set_mdio_data,
>         .get_mdio_data = ravb_get_mdio_data,
> };
> 
> So, I don't think the driver core can call the remove() method for the mii bus
> because it's a part of the ravb driver...

   And because the MDIO module just doesn't have the usual method! :-)
(I meant the EtherAVB driver's remove() method, and that one would be called after
a successful reference count check...)

> By the way, about the mdio-gpio driver, I'm wondering if the mdio-gpio
> driver cannot be removed by rmmod too. (perhaps, we need "rmmod -f" to remove it.)

   You're on your own here. It's fortunate for this patch that I'm not currently loaded
at work! :-)

>>> By the way, I think you have to send this patch to the following maintainers too:
>>> # We can get it by using scripts/get_maintainers.pl.
>>> David S. Miller <davem@...emloft.net> (maintainer:NETWORKING DRIVERS,commit_signer:8/8=100%)
>>> Jakub Kicinski <kuba@...nel.org> (maintainer:NETWORKING DRIVERS)

   Not critical, as DaveM uses the patchwork anyway. He started to be CC'ed on netdev patches
only recently. :-)

[...]

> Best regards,
> Yoshihiro Shimoda

MBR, Sergei