linux-kernel - Re: [PATCH] serial: 8250_lpss: Release Quark MSI vectors on exit

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]

Message-ID: <1483891492.26691.24.camel@linux.intel.com>
Date:   Sun, 08 Jan 2017 18:04:52 +0200
From:   Andy Shevchenko <andriy.shevchenko@...ux.intel.com>
To:     Christoph Hellwig <hch@...radead.org>,
        Andy Shevchenko <andy.shevchenko@...il.com>,
        Tejun Heo <tj@...nel.org>, Bjorn Helgaas <helgaas@...nel.org>
Cc:     Jan Kiszka <jan.kiszka@...mens.com>,
        Christoph Hellwig <hch@....de>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        "linux-serial@...r.kernel.org" <linux-serial@...r.kernel.org>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] serial: 8250_lpss: Release Quark MSI vectors on exit

On Sun, 2017-01-08 at 02:24 -0800, Christoph Hellwig wrote:
> On Thu, Jan 05, 2017 at 12:19:56AM +0200, Andy Shevchenko wrote:
> > NAK, check the PCI devres code, please.

+Cc: Tejun, Bjorn.

> 
> Releasing something through devres that wasn't allocated using a
> devm_* or
> pcim_* function isn't expected, and we should fix that instead.
> pci_free_irq_vectors is _currently_ implemented by calling
> pci_disable_msi and pci_disable_msix, but there is no guarantee for
> that
> in the API.

> Your code works by accident, not by design.  If you want the resources
> to be auto-released you need to add a proper pcim_alloc_irq_vectors
> API.

Though idea sounds sane I disagree this is accidental. The PCI managed
introduction includes among other this one:

+  pcim_enable_device() : after success, all PCI ops become managed

As per commit 9ac7849e35f7 ("devres: device resource management").

Thus, I suppose a new API should follow existing design, or provide a
sane fix. I briefly checked MSI/-X usage and since I'm not so familiar
with the PCI core code, I wouldn't be brave to break it.

Currently I suspect pci_free_irq_vectors() is not friendly with
pcim_release().

-- 
Andy Shevchenko <andriy.shevchenko@...ux.intel.com>
Intel Finland Oy