Re: [PATCH] lpfc: Avoid to disable pci_dev twice

2014-09-09 Thread Mike Qiu
On 08/28/2014 02:34 AM, James Smart wrote: Mike, Can you confirm - the "nulls" this patch correct are because the probe_one and error_detect threads are running concurrently, thus battling ? If so - this fix looks insufficient and we should rework it. Yes, it is. My patch is just a workaro

Re: [PATCH] lpfc: Avoid to disable pci_dev twice

2014-08-27 Thread James Smart
Mike, Can you confirm - the "nulls" this patch correct are because the probe_one and error_detect threads are running concurrently, thus battling ? If so - this fix looks insufficient and we should rework it. Q: why are they allowed to run concurrently ? I could see this solved at the platf

Re: [PATCH] lpfc: Avoid to disable pci_dev twice

2014-07-31 Thread Mike Qiu
On 07/17/2014 02:32 PM, Mike Qiu wrote: Hi, all How about this patch ? Any idea ? In IBM Power servers, when hardware error occurs during probe state, EEH subsystem will call driver's error_detected interface, which will call pci_disable_device(). But driver's probe function also call pci_di

Re: [PATCH] lpfc: Avoid to disable pci_dev twice

2014-07-17 Thread Mike Qiu
On 07/17/2014 10:15 PM, Joe Lawrence wrote: [ +cc linux-pci and Bjorn, comments inline/below ... ] On Thu, 17 Jul 2014 02:32:31 -0400 Mike Qiu wrote: In IBM Power servers, when hardware error occurs during probe state, EEH subsystem will call driver's error_detected interface, which will call

Re: [PATCH] lpfc: Avoid to disable pci_dev twice

2014-07-17 Thread Joe Lawrence
[ +cc linux-pci and Bjorn, comments inline/below ... ] On Thu, 17 Jul 2014 02:32:31 -0400 Mike Qiu wrote: > In IBM Power servers, when hardware error occurs during probe > state, EEH subsystem will call driver's error_detected interface, > which will call pci_disable_device(). But driver's probe

[PATCH] lpfc: Avoid to disable pci_dev twice

2014-07-16 Thread Mike Qiu
In IBM Power servers, when hardware error occurs during probe state, EEH subsystem will call driver's error_detected interface, which will call pci_disable_device(). But driver's probe function also call pci_disable_device() in this situation. So pci_dev will be disabled twice: Device lpfc disabl