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>] [day] [month] [year] [list]
Message-ID: <ed20264b-1fe3-59e2-7c95-d1a224d22b7f@redhat.com>
Date:   Thu, 3 Jun 2021 14:38:50 +1000
From:   Gavin Shan <gshan@...hat.com>
To:     robh@...nel.org
Cc:     Andrew Jones <drjones@...hat.com>, devicetree@...r.kernel.org,
        linux-kernel@...r.kernel.org, Shan Gavin <shan.gavin@...il.com>
Subject: [Question] Format of memory node name

Hi Rob,

Currently, I'm looking into one QEMU bug which is related to FDT memory
nodes. The story is 4 NUMA nodes are specified by the command line, which
is used to start the VM. QEMU needs to create 4 FDT memory nodes for them
and their names are following the format (memory@...t-address) as stated
in the device-tree specification. The 'unit-address' is base address in
'reg' property. Unfortunately, one NUMA node's base address is exactly
same to another one. It means there are two conflicting two memory node
because of their names. It leads to the FDT can't be populated successfully.

For example, the last memory node can't be created because of its conflicting
name in the following scheme.

    NUMA ID       Base address       End address      Memory node name
    -------------------------------------------------------------------
    0             0x00000000         0x20000000       memory@...00000
    1             0x20000000         0x40000000       memory@...00000
    2             0x40000000         0x40000000       memory@...00000
    3             0x40000000         0x40000000       memory@...00000

I'm trying to resolve the issue. There are two options as below. However,
it's not certain the solution will beak the device-tree specification. So
I'm checking with you on this.

(1) Replace the 'unit-address' with NUMA ID
     memory@0
     memory@1
     memory@2
     memory@3

(2) Add suffix to the conflicting memory node names
     memory@...00000
     memory@...00000
     memory@...00000
     memory@...00000-0

Thanks,
Gavin

[link] https://patchwork.kernel.org/project/qemu-devel/patch/20210601073004.106490-1-gshan@redhat.com/

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ