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-next>] [day] [month] [year] [list]
Date:   Sat, 12 Nov 2022 22:41:25 +1100
From:   Andrew Powers-Holmes <aholmes@...om.net>
To:     linux-rockchip@...ts.infradead.org
Cc:     Ondřej Jirman <megi@....cz>,
        Rob Herring <robh+dt@...nel.org>,
        Krzysztof Kozlowski <krzysztof.kozlowski+dt@...aro.org>,
        Heiko Stuebner <heiko@...ech.de>,
        Peter Geis <pgwipeout@...il.com>,
        Frank Wunderlich <frank-w@...lic-files.de>,
        Michael Riesch <michael.riesch@...fvision.net>,
        Yifeng Zhao <yifeng.zhao@...k-chips.com>,
        Sascha Hauer <s.hauer@...gutronix.de>,
        Nicolas Frattaroli <frattaroli.nicolas@...il.com>,
        Chris Morgan <macromorgan@...mail.com>,
        Ezequiel Garcia <ezequiel@...guardiasur.com.ar>,
        Robin Murphy <robin.murphy@....com>,
        Mark Kettenis <mark.kettenis@...all.nl>,
        devicetree@...r.kernel.org, linux-arm-kernel@...ts.infradead.org,
        linux-kernel@...r.kernel.org
Subject: [PATCH 0/1] arm64: dts: rockchip: rk356x: Fix PCIe register and range mappings

The Rockchip RK356x SoCs currently have incorrect, or at least sub-optimal,`reg`
and `ranges` values in their DTS files' PCIe nodes. Ondřej Jirman sent a patch
in [1] to resolve this, but it was not merged due to some issues discovered
during testing (it fixed his issues with devices behind a switch, but broke
directly connected NVMe drives, amongst others - see [2]).

This patch is a reworked of that patch, using the same mappings the Rockchip BSP
kernel uses. Peter sent these up during the discussion in [3] and they've been
tested on his boards as well as Ondřej's, mine, and those of a few others.

Ondřej also sent a patch in [4] with these fixed ranges, but without the fix
for RK3568 as he was not able to test on that SoC. I've included the fixes for
both SoCs as he's happy with that and the patch has not yet been merged.

I have tested these ranges against devices which only map 32-bit ranges, devices
which only map 64-bit, and devices which require both. An Intel i350-T4 NIC does
not enumerate at all with the existing or previous patch's addresses, but works
quite happily with these, as do NVMe drives and every other device I've been
able to test.

MSI/MSI-X has also been tested as working, but does not currently work upstream
due to a workaround needed in the GIC driver which Rockchip are still yet to
issue an erratum for.

Thanks,
Andrew

[1] https://lore.kernel.org/linux-rockchip/20221005085439.740992-1-megi@xff.cz/
[2] https://lore.kernel.org/linux-rockchip/CAMdYzYq3S2rR3Kb61irpV9xHYijNiJY0mkVnJwPrpXzxg_Zh9g@mail.gmail.com/
[3] https://lore.kernel.org/linux-rockchip/CAMdYzYp6ShLqKxdiAjaRFiRF5i+wzfKiQvwPMzyQLAutWZbApg@mail.gmail.com/
[4] https://lore.kernel.org/all/20221107130157.1425882-1-megi@xff.cz/

Andrew Powers-Holmes (1):
  arm64: dts: rockchip: rk356x: Fix PCIe register and range mappings

 arch/arm64/boot/dts/rockchip/rk3568.dtsi | 14 ++++++++------
 arch/arm64/boot/dts/rockchip/rk356x.dtsi |  7 ++++---
 2 files changed, 12 insertions(+), 9 deletions(-)


base-commit: f0c4d9fc9cc9462659728d168387191387e903cc
--
2.38.0

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ