linux-kernel - Re: [PATCH] PCI: shpchp: Fix probing logic inversion

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] [day] [month] [year] [list]

Date:   Mon, 25 Jun 2018 17:18:40 +0100
From:   Marc Zyngier <marc.zyngier@....com>
To:     Bjorn Helgaas <helgaas@...nel.org>,
        Mika Westerberg <mika.westerberg@...ux.intel.com>
Cc:     Bjorn Helgaas <bhelgaas@...gle.com>,
        Lorenzo Pieralisi <lorenzo.pieralisi@....com>,
        linux-pci@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] PCI: shpchp: Fix probing logic inversion

Bjorn, Mika,

On 25/06/18 15:01, Bjorn Helgaas wrote:
> Hi Marc,
> 
> Sorry we broke this!

No worries, gave me a chance to stare at something else! ;-)

> 
> On Thu, Jun 21, 2018 at 05:47:15PM +0100, Marc Zyngier wrote:
>> Until recently, shpc_probe() would bail out pretty early in the
>> absence of the SHPC capability. A logic change in the way the
>> driver now checks that capability makes it go and probe the
>> firmware anyway, with ugly consequences if the system is not
>> ACPI based (my arm64 ThunderX is DT driven, and explodes in
>> a spectacular way after getting a NULL root bridge from the
>> non-existent ACPI tables...).
>>
>> Take this opportunity to move the call to shpchp_is_native()
>> back into shpc_probe(), making it clear that a non-ACPI system
>> is not expected to use this driver.
> 
> But a non-ACPI system *should* be able to use SHPC.

Yeah, I only realized that once Mika pointed it out.

> Here's my understanding of how it should work.  On an ACPI system,
> 
>   - If firmware has _OSC, the OS calls it to request permission to
>     manage the SHPC.  If _OSC grants permission, it should also
>     configure the hardware (interrupts, etc) to give the OS.
>     
>   - If there's no _OSC, shpchp assumes it's allowed to manage the
>     SHPC, and it calls OSHP to configure the hardware appropriately.
> 
> On a non-ACPI system, shpchp assumes there's no firmware involved at
> all, so it can manage the SHPC without doing anything special.
> 
> I see Mika has already posted a patch similar to the first one
> below; I think either of those should fix the problem you're seeing.

Absolutely. With this early exit when root is NULL, my test box is back
up and running.

> The second is an attempt to clean things up so they make a
> little more sense.

I haven't tried that second patch yet, but from reading it, it
definitely helps making sense of this driver.

Thanks,

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