linux-kernel - Re: Boot failure on gru-scarlet-inx with 5.9-rc2

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]

Date:   Mon, 31 Aug 2020 10:27:37 +0100
From:   Marc Zyngier <maz@...nel.org>
To:     Samuel Dionne-Riel <samuel@...nne-riel.com>,
        Rob Herring <robh@...nel.org>
Cc:     devicetree@...r.kernel.org, Frank Rowand <frowand.list@...il.com>,
        linux-kernel@...r.kernel.org
Subject: Re: Boot failure on gru-scarlet-inx with 5.9-rc2

On 2020-08-31 08:18, Samuel Dionne-Riel wrote:
> On Sun, 30 Aug 2020 10:41:42 +0100
> Marc Zyngier <maz@...nel.org> wrote:
> 
> Hi,
> 
>> 
>> Could you try replacing the problematic patch with [1], and let me
>> know whether this changes anything on your end? This patch probably
>> isn't the right approach, but it would certainly help pointing me
>> in the right direction.
>> 
>> [1]
>> https://lore.kernel.org/lkml/20200815125112.462652-2-maz@kernel.org/
> 
> Following through a bisect session to figure out why the Wi-Fi broke
> between 5.8 and 5.9-rc1, I figured out something that you might have in
> mind already.
> 
> It seems that anything that makes of_bus_pci_match return true will
> cause this to happen. This is why your initial fix also fails.
> 
> I believe my understanding is right since applying the following on top
> of 5.9-rc1 also produces the same result.
> 
> --- a/arch/arm64/boot/dts/rockchip/rk3399.dtsi
> +++ b/arch/arm64/boot/dts/rockchip/rk3399.dtsi
> @@ -227,6 +227,7 @@ dmac_peri: dma-controller@...e0000 {
>         };
> 
>         pcie0: pcie@...00000 {
> +               device_type = "pci";
>                 compatible = "rockchip,rk3399-pcie";
>                 reg = <0x0 0xf8000000 0x0 0x2000000>,
>                       <0x0 0xfd000000 0x0 0x1000000>;
> 
> 
> This was found out since the Wi-Fi pci-based ath10k Wi-Fi broke, with
> 2f96593ecc37e98bf99525f0629128080533867f, which changes stuff around
> pci bus... things...
> 
> Am I understanding right that your fix(es) were related to the change
> set where the commit is found?
> 
> My intuition is that the commit causing the boot issue could be related
> to changes with PCI or PCIe subsystems, and that your fix for
> of_bus_pci_match is a red herring, that only surfaced the existing
> issue.
> 
> This is backed by applying the previous dts patch on top of 2f96593e,
> and having Wi-Fi work. I would assume that between that commit and
> 5.9-rc1 there is a commit that causes the complete failure to boot,
> which is unrelated to the first identified commit on 5.9-rc2.

Ah, so actually anything that *enables pcie* kills your system.
Great investigative work!

> 
> And backed by a further bisection with this that points to
> d84c572de1a360501d2e439ac632126f5facf59d being the actual change that
> causes the tablet to fail to boot, as long as the pcie0 node is
> identified as pci properly.
> 
> I am unsure if I should add as a Cc everyone involved in that change
> set, though the author (coincidentally) is already in the original list
> of recipients.

I've deliberately moved Rob from Cc to To... ;-)

> Any additional thoughts from this additional information?

What you could do is to start looking at which of the pci_is_root_bus()
changes breaks PCIe on this system.  The fact that it breaks on your
system and not on mine is a bit puzzling.

Thanks,

         M.
-- 
Jazz is not dead. It just smells funny...