Re: Broken pci_block_user_cfg_access interface

[Jan Kiszka <jan.kiszka@siemens.com>]

On 2011-08-29 18:23, Michael S. Tsirkin wrote:
> On Mon, Aug 29, 2011 at 06:14:39PM +0200, Jan Kiszka wrote:
>> On 2011-08-29 17:58, Michael S. Tsirkin wrote:
>>> On Mon, Aug 29, 2011 at 05:42:16PM +0200, Jan Kiszka wrote:
>>>> I still don't get what prevents converting ipr to allow plain mutex
>>>> synchronization. My vision is:
>>>>  - push reset-on-error of ipr into workqueue (or threaded IRQ?)
>>>>  - require mutex synchronization for common config space access
>>>
>>> Meaning pci_user_ read/write config?
>>
>> And pci_dev_reset, yes.
>>
>>>
>>>>     and the
>>>>    full reset cycle
>>>>  - only exception: INTx status/masking access
>>>>     => use pci_lock + test for reset_in_progress, skip operation if
>>>>        that is the case
>>>>
>>>> That would allow to drop the whole block_user_cfg infrastructure.
>>>>
>>>> Jan
>>>
>>> We still need to block userspace access while INTx does
>>> the status/masking access, right?
>>
>> Yes, pci_lock would do that for us.
> 
> Well this means block_user_cfg is not going away,
> this is what it really is: pci_lock + a bit to lock out userspace.

I does as we only end up with a mutex and pci_lock. No more hand-crafted
queuing/blocking/waking.

INTx masking is a bit special as it's the only thing that truly requires
atomic context. But that's something we should address generically anyway.

Jan

-- 
Siemens AG, Corporate Technology, CT T DE IT 1
Corporate Competence Center Embedded Linux