Closed agherzan closed 8 years ago
This issue is caused by re-enable duplicate APIC IRQ number which is sharing same irqdomain with SDHCI PCI, so association with this irqdomain has already been created, not necessary to associate again. That's not big issue for normal running. Close it now.
So the resolution is won't fix
?
Look into upstream solution for this issue as well: http://git.kernel.org/cgit/linux/kernel/git/tip/tip.git/commit/?h=x86/platform&id=bb27570525a71f48347ed0e0c265063e7952bb61
Yes, I was aware of this workaround. But don't understand why was this issue closed.
I have no idea.
Very thankful for Linux member's support. Very welcome to author and introduction of appliable patches from Linux Community. Sorry to my last brief comments.
First of all, the patch mentioned is good for workaround, though not very compatible with our codebase(Linux 3.19.5). In the codebase before this patch from Andy-shev's, it will fail to enable other PCI devices using IRQ0, that was a serious problem on a BIOS platform and need Message Bus work.
But in our codebase, it processes OK, no reaction of fail path as return, probe&init process works OK, just report a warn to avoid repeated association and continue the process. And we could let it be ease because it is not from MP interrupt source record, so this device is still wokring in our codebase, only cause one newed IRQ description struct alone.. and our platform don't need BIOS and MBI function work,
And like the description of this issue, currently no related functional or side-effect problems have reported, and I don't have enough time to dig deeply enough to kill it from the root for this kind of warning issues, so I recommend ignore it until if a new opened issue could be related to this issue.
Basically the issue is in firmware which assigns resources (IRQ) wrongly in PCI config space. PCI specification doesn't allow interrupt to be enabled and be 0. So, eMMC0 shouldn't use IRQ0 in the first place. Second issue is so called "global configuration space for HS UART" which by some reason is represented as PCI device (00:04.0).
P.S. I'm sorry I didn't get how mentioned patch from the upstream makes MailBox not working. MBI, as far as I know, doesn't require any interrupt and thus just work w/o it. Ah, you don't use it at all.
Using current head of 3.19 WIP branch I always get the following warning when booting my edison:
Regards, Andrei