linux-kernel - Re: [PATCH v1 4/4] PCI: Allow extend_bridge_window() to shrink resource if necessary

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:   Wed, 15 Jan 2020 14:51:02 +0000
From:   Nicholas Johnson <nicholas.johnson-opensource@...look.com.au>
To:     Mika Westerberg <mika.westerberg@...ux.intel.com>
CC:     Bjorn Helgaas <helgaas@...nel.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        "linux-pci@...r.kernel.org" <linux-pci@...r.kernel.org>,
        Benjamin Herrenschmidt <benh@...nel.crashing.org>,
        Logan Gunthorpe <logang@...tatee.com>
Subject: Re: [PATCH v1 4/4] PCI: Allow extend_bridge_window() to shrink
 resource if necessary

On Mon, Jan 13, 2020 at 06:21:50PM +0200, Mika Westerberg wrote:
> On Wed, Jan 08, 2020 at 01:36:04AM +0000, Nicholas Johnson wrote:
> > > Where's the patch that changes the caller so "new_size" may be smaller
> > > than "size"?  I guess it must be "[3/3] PCI: Consider alignment of
> > > hot-added bridges ..." because that's the only one that makes a
> > > non-trivial change, right?
> > 
> > As above, there was always a possibility of the new_size being smaller. 
> > For some reason, 1M is assigned to bridges, even if nothing is below 
> > them (for example, unused non hotplug bridges in a Thunderbolt dock). It 
> > may be an edge case if we are low on space, but theoretically it can 
> > happen.
> > 
> > Also, when writing this, Mika was not interested in using hpmemsize, 
> > which, when used, will cause new_size to be smaller than the current 
> > size (actual size and add_size combined).
> 
> Just a small correction here about my motivation. So I'm testing on a
> hardware where the BIOS assigns initial resources to the root/downstream
> ports which is the majority of Thunderbolt capable PC systems nowadays.
> Therefore the user does not need to pass any additional command line
> parameters to get the ports working properly.
> 
> However, I'm of course interested in getting Linux PCI resource
> management as good as possible regardless whether the firmware/BIOS
> assigns them or not ;-)
Sorry, I was not meant to say you were not interested in getting it as 
good as possible. At the time, you had a goal to achieve (which you did) 
and at that point in time, it would not have been feasible to use 
pci=hpmemsize or similar before my patches were applied:

  c13704f5685d ("PCI: Avoid double hpmemsize MMIO window assignment")
  d7b8a217521c ("PCI: Add "pci=hpmmiosize" and "pci=hpmmioprefsize" parameters")

What I was trying to say was not that you were not interested, but more 
that it was not a primary motivation for you at the time. Does this 
sound more accurate? Poor wording on my behalf.