linux-kernel - Re: [PATCH V4 4/6] PCI: tegra: Continue unconfig sequence even if parts fail

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: <20201201144420.GA4397@e121166-lin.cambridge.arm.com>
Date:   Tue, 1 Dec 2020 14:44:20 +0000
From:   Lorenzo Pieralisi <lorenzo.pieralisi@....com>
To:     Thierry Reding <thierry.reding@...il.com>
Cc:     Vidya Sagar <vidyas@...dia.com>, robh+dt@...nel.org,
        bhelgaas@...gle.com, jonathanh@...dia.com,
        amanharitsh123@...il.com, dinghao.liu@....edu.cn, kw@...ux.com,
        linux-pci@...r.kernel.org, linux-tegra@...r.kernel.org,
        linux-kernel@...r.kernel.org, kthota@...dia.com,
        mmaddireddy@...dia.com, sagar.tv@...il.com
Subject: Re: [PATCH V4 4/6] PCI: tegra: Continue unconfig sequence even if
 parts fail

On Tue, Dec 01, 2020 at 03:24:24PM +0100, Thierry Reding wrote:
> On Mon, Nov 30, 2020 at 12:10:07PM +0000, Lorenzo Pieralisi wrote:
> > On Mon, Nov 09, 2020 at 10:49:35PM +0530, Vidya Sagar wrote:
> > > Currently the driver checks for error value of different APIs during the
> > > uninitialization sequence. It just returns from there if there is any error
> > > observed for one of those calls. Comparatively it is better to continue the
> > > uninitialization sequence irrespective of whether some of them are
> > > returning error. That way, it is more closer to complete uninitialization.
> > 
> > Hi Vidya, Thierry,
> > 
> > I can apply this series (dropping patches as suggested by Thierry),
> > before though I wanted to ask you if this patch is really an
> > improvement, it is hard to understand why skipping some error
> > codes is OK for device correct operations to continue, maybe it
> > is worth describing why some of those failures aren't really
> > fatal.
> > 
> > Please let me know, thanks.
> 
> As explained in the commit message, the idea is to continue tearing
> down even if things fail somewhere in the middle, because that ensures
> that the hardware gets as close to an "uninitialized" state as possible.
> If for example the first reset assert were to fail, then none of the
> PHYs get disabled, the regulator stays on and the clocks stays on, all
> of which can continue draining power after the controller has already
> been torn down.

Understood. By reading the code it looked weird that eg a reset failure
was tolerable - I thought an error would be fatal (I don't know what are
the consequences for instance on a subsequent resume) but it looks like
it actually is not, that's the only point I raised.

> So yes, I think this is an improvement. It's unclear to me what you're
> asking for, though. Would you rather have a comment somewhere near the
> tegra_pcie_unconfig_controller() function that states the same thing as
> the commit message?

It could be useful but it is up to you, I will merge the series as-is
or I can add it myself, as you wish.

Thanks,
Lorenzo