Bug 217680

Summary:	PCI: Loongson: 8b3517f88ff2 breaks PCIe on some system
Product:	Drivers	Reporter:	jiaxun.yang
Component:	PCI	Assignee:	drivers_pci (drivers_pci)
Status:	NEW ---
Severity:	normal	CC:	bagasdotme, carnil, jiaxun.yang
Priority:	P3
Hardware:	Mips64
OS:	Linux
Kernel Version:		Subsystem:
Regression:	Yes	Bisected commit-id:	8b3517f88ff2983f52698893519227c10aac90b2/

Description jiaxun.yang 2023-07-18 13:07:32 UTC

After 8b3517f88ff2 ("PCI: loongson: Prevent LS7A MRRS increases") backported to stable, we have received multiple regression reports including one from Debian build farm[1] that some systems failed to initialize some PCIe devices.

I was able to reproduce on one of my Loongson-3A4000+7A system. It seems like the root cause is firmware on those systems didn't set MRRS properly but 8b3517f88ff2 ("PCI: loongson: Prevent LS7A MRRS increases") assumed firmware's maximum MRRS is always valid.

[1]: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1035587

Comment 1 jiaxun.yang 2023-07-19 08:32:07 UTC

Mark as regression.

Comment 2 Bagas Sanjaya 2023-07-23 01:23:51 UTC

(In reply to jiaxun.yang from comment #0)
> After 8b3517f88ff2 ("PCI: loongson: Prevent LS7A MRRS increases") backported
> to stable, we have received multiple regression reports including one from
> Debian build farm[1] that some systems failed to initialize some PCIe
> devices.
> 
> I was able to reproduce on one of my Loongson-3A4000+7A system. It seems
> like the root cause is firmware on those systems didn't set MRRS properly
> but 8b3517f88ff2 ("PCI: loongson: Prevent LS7A MRRS increases") assumed
> firmware's maximum MRRS is always valid.
> 
> [1]: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1035587

I see that you're about to post your patch that (supposedly) fixes
this regression here [1] . Where is one?

[1]: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1035587#55