general issue with suspend/resume with iwn(4)/bge(4)

[Serge Semenenko]

John Baldwin wrote:
> On Monday 09 November 2009 1:14:18 pm Serge Semenenko wrote:
>   
>>  John Baldwin wrote: 
>> On Sunday 08 November 2009 6:19:09 am Bernhard Schmidt wrote:
>>   
>> Hi,
>>
>> I hope this is the correct list for an issue like that, if not, a pointer 
>> would be appreciated.
>>
>> I've been in contact with Mykola Dzham quite some time now and we are trying 
>> to figure out a resume issue on his iwn(4) device. It does seem that this 
>> device does not come up correctly after suspend. The interesting part is, 
>>     
> that 
>   
>> even pciconf -l -bcv ist not able to get all information.
>>
>> Before suspend:
>> iwn0 at pci0:6:0:0:        class=0x028000 card=0x13018086 chip=0x42328086 
>> rev=0x00 hdr=0x00
>>     vendor     = 'Intel Corporation'
>>     device     = 'Carte Intel WiFi Link 5100 AGN (Intel WiFi Link 5100)'
>>     class      = network
>>     bar   [10] = type Memory, range 64, base 0xec800000, size 8192, enabled
>>     cap 01[c8] = powerspec 3  supports D0 D3  current D0
>>     cap 05[d0] = MSI supports 1 message, 64 bit enabled with 1 message
>>     cap 10[e0] = PCI-Express 1 endpoint max data 128(128) link x1(x1)
>>
>> After resume:
>> iwn0 at pci0:6:0:0:        class=0x028000 card=0x13018086 chip=0x42328086 
>> rev=0x00 hdr=0x00
>>     vendor     = 'Intel Corporation'
>>     device     = 'Carte Intel WiFi Link 5100 AGN (Intel WiFi Link 5100)'
>>     class      = network
>>     
>>
>> Are you sure you didn't forget the extra options to pciconf here?  The bar
>> should definitely not disappear since we save that state in software, not
>> in hardware.  Also, the capability pointer register is set by the hardware,
>> software never changes it.
>>
>>   
>>
>>  It looks similar to PR http://www.freebsd.org/cgi/query-pr.cgi?pr=135070 
>>     
> for me. And if I understood right you're already working on the solution...
>
> No, having the capability registers and a BAR disappear after they were 
> programmed is entirely different.  That PR is about being able to allocate 
> space for the BAR on boot, not about losing it entirely after resume.
>
>   
Not sure about nature of things happened but on my system resources 
programmed with mentioned in PR hack are also disappears on resume and 
to get things working the hack should be applied both on attach and on 
resume.