Yinghai Lu wrote:
Kenji Kaneshige wrote:
Hi Yinghai,
I would like to reconfirm what is the problem you're trying to solve
before reviewing and testing the additional patch. To be honest, your
patch looks more and more complicated and it is becoming difficult
for me to review and test it.
the real problem:
1. boot time:
BIOS separate IO range between several IOHs, and on some slots, BIOS assign the resource to the bridge, but stop
assigning resource to the device under that bridge, because the device need big resource.
so patch1 is trying to
a. pci assign unassign and record the failed device resource.
b. clear the BIOS assigned resource of the parent bridge of fail device
c. go back and call pci assign unsigned
d. if it still fail, will go up more bridges. and clear and try again.
2. hotplug:
BIOS separate IO range between several IOHs, and on some slots, BIOS assign the resource to every bridge. (8M)
but when insert one card that big resource, the card can not get resource. because kernel will not touch the
bridge resource.
so patch2 is trying to
a. assign resource to devices with that slot. and record fail devices
b. if there is some failed, will clear sepcifically io port of bridge, or mmio of bridge, or mmio pref of bridge.
c. try to assign the parent bridge of the slot.
so it will keep original assigned resource by BIOS if possible.
and you have tested patch1 and patch2 in V11, but said patch1 may have shrinking resource problem.
the patch3 is addressing the patch1 that could shrinking resource for non-pcie hotplug bridge...
Thank you for the explanation. The patch3 seems to solve my concern.
Your patch only touch the leaf bridge at the 2nd try of resource
assignment. IIRC, this behavior is to prevent from shrinking bridge
resources. Am I correct? I'm not sure but I think we don't need this
behavior because now that we have another mechanism to prevent
from shrinking bridge resource.
Thanks,
Kenji Kaneshige
By the way, if your problem is that BIOS doesn't assign the resource
to the parent bridge (root port or switch downstream port) of PCIe
hotplug slots, I guess we can improve it with simpler way. I made a
sample patches (no enough testing). Please take a look. Patches are
- [PATCH 1/2] pciehp: remove redundancy in bridge resource allocation
- [PATCH 2/2] pciehp: add support for bridge resource reallocation
like some earlier version of patch2 (release it them all at first) but have pciehp_realloc parameter.
could be useful in some case when current patch2 (try and increase) could use up all space.
i like to have that after patch2.
YH
--
To unsubscribe from this list: send the line "unsubscribe linux-pci" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
