| 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
| ||
|
Message-Id: <20231002120511.594737-1-suijingfeng@loongson.cn> Date: Mon, 2 Oct 2023 20:05:09 +0800 From: Sui Jingfeng <suijingfeng@...ngson.cn> To: Bjorn Helgaas <bhelgaas@...gle.com> Cc: linux-pci@...r.kernel.org, linux-kernel@...r.kernel.org Subject: [PATCH pci-next v6 0/2] PCI/VGA: Make the vga_is_firmware_default() less arch-dependent The vga_is_firmware_default() only works on x86 and ia64 currently, it is a no-op on the rest of the architectures. This patch completes the implementation for it, the added code tries to capture the PCI (e) VGA device that owns the firmware framebuffer before the PCI resource relocation. Since only one GPU could owns the firmware fb in normal case, things are almost done once we have determined the boot VGA device successfully. Note that this patch requires the target platform has a way to set up the kernel's screen_info. On machines with muiltiple GPU co-exist and On UEFI environment, the firmware framebuffer should be put into the VRAM BAR of the primary GPU. For the arbitration purpose itself, changing PCI class code of the GPU to be non-primary is not required. The provided method is tested on x86, arm64 and loongarch, all of the machine with UEFI firmware shipped. See below for more information. 1) LS3A5000+LS7A1000 platform with three video cards: $ lspci | grep VGA 00:06.1 VGA compatible controller: Loongson Technology LLC DC (Display Controller) (rev 01) 03:00.0 VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Caicos XT [Radeon HD 7470/8470 / R5 235/310 OEM] 07:00.0 VGA compatible controller: S3 Graphics Ltd. Device 9070 (rev 01) 08:00.0 VGA compatible controller: S3 Graphics Ltd. Device 9070 (rev 01) Before apply this series: pci 0000:00:06.1: vgaarb: setting as boot VGA device pci 0000:00:06.1: vgaarb: VGA device added: decodes=io+mem,owns=io+mem,locks=none pci 0000:03:00.0: vgaarb: VGA device added: decodes=io+mem,owns=none,locks=none pci 0000:07:00.0: vgaarb: VGA device added: decodes=io+mem,owns=none,locks=none pci 0000:08:00.0: vgaarb: VGA device added: decodes=io+mem,owns=none,locks=none vgaarb: loaded After apply this series: pci 0000:03:00.0: vgaarb: BAR 0: [mem 0xe0050000000-0xe005fffffff 64bit pref] contains firmware FB [0xe0050000000-0xe00500ea5ff] pci 0000:00:06.1: vgaarb: setting as boot VGA device pci 0000:03:00.0: vgaarb: setting as boot VGA device (overriding previous) pci 0000:03:00.0: vgaarb: VGA device added: decodes=io+mem,owns=none,locks=none pci 0000:07:00.0: vgaarb: VGA device added: decodes=io+mem,owns=none,locks=none pci 0000:08:00.0: vgaarb: VGA device added: decodes=io+mem,owns=none,locks=none vgaarb: loaded $ dmesg | grep 0000:03:00.0 pci 0000:03:00.0: [1002:6778] type 00 class 0x030000 pci 0000:03:00.0: reg 0x10: [mem 0xe0050000000-0xe005fffffff 64bit pref] pci 0000:03:00.0: reg 0x18: [mem 0xe0065300000-0xe006531ffff 64bit] pci 0000:03:00.0: reg 0x20: [io 0x20000-0x200ff] pci 0000:03:00.0: reg 0x30: [mem 0xfffe0000-0xffffffff pref] pci 0000:03:00.0: vgaarb: BAR 0: [mem 0xe0050000000-0xe005fffffff 64bit pref] contains firmware FB [0xe0050000000-0xe00500ea5ff] pci 0000:03:00.0: BAR 0: assigned [mem 0xe0030000000-0xe003fffffff 64bit pref] pci 0000:03:00.0: BAR 2: assigned [mem 0xe0065200000-0xe006521ffff 64bit] pci 0000:03:00.0: BAR 6: assigned [mem 0xe0065220000-0xe006523ffff pref] pci 0000:03:00.0: BAR 4: assigned [io 0x5000-0x50ff] pci 0000:03:00.0: vgaarb: setting as boot VGA device (overriding previous) pci 0000:03:00.0: vgaarb: VGA device added: decodes=io+mem,owns=none,locks=none Loongson UEFI firmware does not support specify which GPU to be the primary, the firmware set the ATI GPU(03:00.0) as the primary GPU with this hardware configuration by hardcode. The problem is that VGAARB can not override the platform integrated one(00:06.1) before apply this series. Please note that BAR 0 of the ATI GPU moved by PCI core from [0xe0050000000-0xe005fffffff] to [0xe0030000000-0xe003fffffff], the vga_is_firmware_default() function will return wrong results by simply remove #ifdefs while without take relocation into account. 2) ARM64 (Kunpeng 920) with three video card: Before apply this series: pci 0000:02:00.0: vgaarb: setting as boot VGA device pci 0000:02:00.0: vgaarb: VGA device added: decodes=io+mem,owns=none,locks=none pci 0000:05:00.0: vgaarb: setting as boot VGA device (overriding previous) <--- (Because it has IO or MEM enabled) pci 0000:05:00.0: vgaarb: VGA device added: decodes=io+mem,owns=none,locks=none pci 0000:06:00.0: vgaarb: VGA device added: decodes=io+mem,owns=none,locks=none vgaarb: loaded After apply this series: pci 0000:05:00.0: vgaarb: BAR 0: [mem 0x80010000000-0x8001fffffff 64bit pref] contains firmware FB [0x80010000000-0x800101e77ff] pci 0000:02:00.0: vgaarb: setting as boot VGA device pci 0000:02:00.0: vgaarb: VGA device added: decodes=io+mem,owns=none,locks=none pci 0000:05:00.0: vgaarb: Boot VGA selected by firmware pci 0000:05:00.0: vgaarb: setting as boot VGA device (overriding previous) <--- (Because it owns firmware framebuffer) pci 0000:05:00.0: vgaarb: VGA device added: decodes=io+mem,owns=none,locks=none pci 0000:06:00.0: vgaarb: VGA device added: decodes=io+mem,owns=none,locks=none vgaarb: loaded 3) x86 with three video card lspci | grep VGA 05:00.0 VGA compatible controller: Silicon Motion, Inc. SM750 (rev a1) 0c:00.0 VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Navi 10 [Radeon RX 5600 OEM/5600 XT / 5700/5700 XT] (rev c1) 0d:00.0 VGA compatible controller: Jingjia Microelectronics Co Ltd Device 9100 (rev 01) Before apply this series: pci 0000:05:00.0: vgaarb: setting as boot VGA device pci 0000:05:00.0: vgaarb: VGA device added: decodes=io+mem,owns=none,locks=none pci 0000:0c:00.0: vgaarb: setting as boot VGA device (overriding previous) pci 0000:0c:00.0: vgaarb: VGA device added: decodes=io+mem,owns=io+mem,locks=none pci 0000:0d:00.0: vgaarb: VGA device added: decodes=io+mem,owns=none,locks=none vgaarb: loaded amdgpu 0000:0c:00.0: vgaarb: deactivate vga console amdgpu 0000:0c:00.0: vgaarb: changed VGA decodes: olddecodes=io+mem,decodes=none:owns=io+me After apply this series: pci 0000:0c:00.0: vgaarb: BAR 0: [mem 0xa0000000-0xafffffff 64bit pref] contains firmware FB [0xa0000000-0xa1fa3fff] pci 0000:05:00.0: vgaarb: setting as boot VGA device pci 0000:05:00.0: vgaarb: VGA device added: decodes=io+mem,owns=none,locks=none pci 0000:0c:00.0: vgaarb: Boot VGA selected by firmware pci 0000:0c:00.0: vgaarb: setting as boot VGA device (overriding previous) pci 0000:0c:00.0: vgaarb: VGA device added: decodes=io+mem,owns=io+mem,locks=none pci 0000:0d:00.0: vgaarb: VGA device added: decodes=io+mem,owns=none,locks=none vgaarb: loaded amdgpu 0000:0c:00.0: vgaarb: deactivate vga console amdgpu 0000:0c:00.0: vgaarb: changed VGA decodes: olddecodes=io+mem,decodes=none:owns=io+mem v2: * Fix test robot warnnings and fix typos v3: * Fix linkage problems if the global screen_info is not exported v4: * Handle linkage problems by hiding behind of CONFIG_SYSFB, * Drop side-effects and simplify. v5: * Print the BAR and the framebuffer region (Bjorn) * Use pci_dev_for_each_resource() (Bjorn) * Cleanup the old mechanisms (Bjorn) * Make the commit log simple by moving the extraneous details to cover letter (Bjorn) * Carry on test on arm64 v6: * Resolve merge conflect on pci-next branch. Sui Jingfeng (2): PCI/VGA: Make the vga_is_firmware_default() less arch-dependent PCI/VGA: Remove vga_is_firmware_default() function drivers/pci/vgaarb.c | 116 ++++++++++++++++++++++++++++--------------- 1 file changed, 76 insertions(+), 40 deletions(-) base-commit: 87ab156b7bb2a117a79692a96239c667f5b4f08e -- 2.34.1
Powered by blists - more mailing lists