Hi, Thorsten here, the Linux kernel's regression tracker. Top-posting for once, to make this easily accessible to everyone.
Just wondering what's up here. The patch this thread is about afaics was supposed to fix a regression reported in July (https://bugzilla.kernel.org/show_bug.cgi?id=217680 ), but has made not steps closer to get mainlined during the past few weeks. Is there a reason, or did it maybe fell through the cracks?
Jiaxun Yang, from it quick look it seems like you wanted to post a v3, but never did so; but I might be mistaken there.
Ciao, Thorsten (wearing his 'the Linux kernel's regression tracker' hat) -- Everything you wanna know about Linux kernel regression tracking: https://linux-regtracking.leemhuis.info/about/#tldr If I did something stupid, please tell me, as explained on that page.
#regzbot poke
On 20.09.23 14:33, Linux regression tracking (Thorsten Leemhuis) wrote:
[CCing the regression list, as it should be in the loop for regressions: https://docs.kernel.org/admin-guide/reporting-regressions.html]
On 07.09.23 07:08, Manivannan Sadhasivam wrote:
On Thu, Sep 07, 2023 at 11:13:00AM +0800, Jiaxun Yang wrote:
在 2023/9/7 9:18, Manivannan Sadhasivam 写道: [...]
Why do you need to walk through every single device instead of just bridges? I'm not the maintainer, but my suggestion is to go for Huacai Chen's solution.
Thanks for your reply, unfortunately Huacai's solution is impractical in this case.
The problem we have, is firmware (or BIOS) setting improper MRRS for devices attached under those bridges. So we have to fix up MRRS for every single device. We can't iterate child device in bridge quirk because there is no guarantee that bridge will be probed before it's child device, partly due to hotplug.
Okay, this clarifies and also warrants improvement in commit message.
You could also use pci_walk_bus() after pci_host_probe() to iterate over the child devices under root bridge and set MRRS. IMO that would look neat.
Hi, Thorsten here, the Linux kernel's regression tracker. What's the status here? The regression that was supposed to be fixed by the patched that started this thread was reported 9 weeks ago[1] and the culprit made it to many stable kernels as well. Would be really good to finally fix this, as a regression like this should ideally be fixed within 2 to 3 weeks (in both mainline and stable). With a revert if necessary -- is this maybe still a option, or would that cause more trouble then it solved (I guess that's the case).
[1] https://bugzilla.kernel.org/show_bug.cgi?id=217680
Ciao, Thorsten (wearing his 'the Linux kernel's regression tracker' hat)
Everything you wanna know about Linux kernel regression tracking: https://linux-regtracking.leemhuis.info/about/#tldr If I did something stupid, please tell me, as explained on that page.
This quirk has been in tree for a while, until Huacai refactored it and broke some systems in 8b3517f88ff2 ("PCI: loongson: Prevent LS7A MRRS increases").
Also to note that ks_pcie_quirk in drivers/pci/controller/dwc/pci-keystone.c uses similar approach.
This avoids iterating over bridges/devices two times.
Also, please rename firmware to BIOS, as firmware commonly represents the software running on PCIe endpoint devices.
Ack, will fix in next reversion.
Thanks
- Jiaxun
- Mani
[...]