netdev - Re: RTL8402 stops working after hibernate/resume

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: <1c2d888a-5702-cca9-195c-23c3d0d936b9@redhat.com>
Date:   Wed, 30 Sep 2020 11:04:08 +0200
From:   Hans de Goede <hdegoede@...hat.com>
To:     Heiner Kallweit <hkallweit1@...il.com>,
        Petr Tesarik <ptesarik@...e.cz>
Cc:     Realtek linux nic maintainers <nic_swsd@...ltek.com>,
        netdev@...r.kernel.org, linux-clk@...r.kernel.org
Subject: Re: RTL8402 stops working after hibernate/resume

Hi,

On 9/29/20 10:35 PM, Heiner Kallweit wrote:
> On 29.09.2020 22:08, Hans de Goede wrote:

<snip>

>> Also some remarks about this while I'm being a bit grumpy about
>> all this anyways (sorry):
>>
>> 1. 9f0b54cd167219 ("r8169: move switching optional clock on/off
>> to pll power functions") commit's message does not seem to really
>> explain why this change was made...
>>
>> 2. If a git blame would have been done to find the commit adding
>> the clk support: commit c2f6f3ee7f22 ("r8169: Get and enable optional ether_clk clock")
>> then you could have known that the clk in question is an external
>> clock for the entire chip, the commit message pretty clearly states
>> this (although "the entire" part is implied only) :
>>
>> "On some boards a platform clock is used as clock for the r8169 chip,
>> this commit adds support for getting and enabling this clock (assuming
>> it has an "ether_clk" alias set on it).
>>
> Even if the missing clock would stop the network chip completely,
> this shouldn't freeze the system as described by Petr.
> In some old RTL8169S spec an external 25MHz clock is mentioned,
> what matches the MII bus frequency. Therefore I'm not 100% convinced
> that the clock is needed for basic chip operation, but due to
> Realtek not releasing datasheets I can't verify this.

Well if that 25 MHz is the only clock the chip has, then it basically
has to be on all the time since all clocked digital ASICs cannot work
without a clock.  Now pci-e is a packet-switched point-to-point bus,
so the ethernet chip not working should not freeze the entire system,
but I'm not really surprised that even though it should not do that,
that it still does.

> But yes, if reverting this change avoids the issue on Petr's system,
> then we should do it. A simple mechanical revert wouldn't work because
> source file structure has changed since then, so I'll prepare a patch
> that effectively reverts the change.

Great, thank you.

Regards,

Hans