[OpenWrt-Devel] AR9344 pcie fails to start (sometimes)

[John Crispin]

On 10/12/2014 14:07, Helmut Schaa wrote:
> On Wed, Dec 10, 2014 at 1:54 PM, John Crispin <blogic at openwrt.org>
> wrote:
>> 
>> 
>> On 10/12/2014 13:43, Helmut Schaa wrote:
>>> Hi,
>>> 
>>> Did anyone else see issues with ar9344 PCIe not coming up
>>> sometimes (it is really rare in my case)?
>>> 
>>> Good boot dmesg:
>>> 
>>> [    0.070000] registering PCI controller with io_map_base
>>> unset [    0.280000] bio: create slab <bio-0> at 0 [
>>> 0.280000] PCI host bridge to bus 0000:00 [    0.280000] pci_bus
>>> 0000:00: root bus resource [mem 0x10000000-0x13ffffff] [
>>> 0.280000] pci_bus 0000:00: root bus resource [io  0x0000] [
>>> 0.280000] pci_bus 0000:00: No busn resource found for root
>>> bus, will use [bus 00-ff] [    0.280000] pci 0000:00:00.0:
>>> [168c:0030] type 00 class 0x028000 [    0.280000] pci
>>> 0000:00:00.0: invalid calibration data [    0.290000] pci
>>> 0000:00:00.0: reg 10: [mem 0x00000000-0x0001ffff 64bit] [
>>> 0.290000] pci 0000:00:00.0: reg 30: [mem 0x00000000-0x0000ffff 
>>> pref] [    0.290000] pci 0000:00:00.0: supports D1 [
>>> 0.290000] pci 0000:00:00.0: PME# supported from D0 D1 D3hot [
>>> 0.290000] pci_bus 0000:00: busn_res: [bus 00-ff] end is updated
>>> to 00 [    0.290000] pci 0000:00:00.0: BAR 0: assigned [mem 
>>> 0x10000000-0x1001ffff 64bit] [    0.290000] pci 0000:00:00.0:
>>> BAR 6: assigned [mem 0x10020000-0x1002ffff pref] [    0.290000]
>>> pci 0000:00:00.0: using irq 40 for pin 1
>>> 
>>> Bad boot dmesg:
>>> 
>>> [    0.070000] ar724x-pci ar724x-pci: PCIe link is down [
>>> 0.070000] registering PCI controller with io_map_base unset [
>>> 0.280000] bio: create slab <bio-0> at 0 [    0.280000] PCI host
>>> bridge to bus 0000:00 [    0.280000] pci_bus 0000:00: root bus
>>> resource [mem 0x10000000-0x13ffffff] [    0.280000] pci_bus
>>> 0000:00: root bus resource [io  0x0000] [    0.280000] pci_bus
>>> 0000:00: No busn resource found for root bus, will use [bus
>>> 00-ff] [    0.280000] pci_bus 0000:00: busn_res: [bus 00-ff]
>>> end is updated to 00
>>> 
>>> In this state no device is detected on the PCI bus.
>>> 
>>> Any ideas?
>>> 
>> 
>> we had the same bug on mtk targets recently to fix it we did
> 
> Thanks, do you have a link to the appropriate commit somewhere?
> 

https://dev.openwrt.org/browser/trunk/target/linux/ramips/patches-3.14/999-pci-reset.patch

and

https://dev.openwrt.org/changeset/43292

>> * made sure the /PCIE_RST goes high before pulling it low and
>> then up again so that it gets a proper toggle and not just half
>> of one. * put the pci core into reset before rebooting the
>> system
>> 
>> i would start by testing the first one to see if the pcie device
>> gets a proper card reset upon boot up and also check the timings.
>> my tests showed that 50ms is the minimum required. however that
>> number is just trial and error so it might be SoC or card
>> specific
> 
> I'll give it a try but it really happens rarely, so lots of boot 
> cycles to reproduce.

trigger a ssid scan and trhen reboot while the call is active to make
sure the cards firmware/microcode is busy. this allowed us to reliably
trigger the case where it got stuck




> 
> Helmut _______________________________________________ 
> openwrt-devel mailing list openwrt-devel at lists.openwrt.org 
> https://lists.openwrt.org/cgi-bin/mailman/listinfo/openwrt-devel
> 
_______________________________________________
openwrt-devel mailing list
openwrt-devel at lists.openwrt.org
https://lists.openwrt.org/cgi-bin/mailman/listinfo/openwrt-devel