Messages in this thread | | | Date | Sat, 10 Apr 2004 18:36:38 -0700 | From | Pete Zaitcev <> | Subject | Re: Patch for usb-storage in 2.4 [linux-usb-devel] |
| |
On Sat, 10 Apr 2004 17:09:57 -0700 Matthew Dharm <mdharm-kernel@one-eyed-alien.net> wrote:
> I'm uncertain if your handling of the io_request_lock is right.... but > getting information on how to handle that has been like pulling teeth, so > I'm inclined to trust your wide-scale testing on this.
It's not wide-scale, unfortunately, that's why I need more review and testing of it. Those "Enterprise" people are mostly interested in very specific things, in particular Bladecenter and JS-20, Dell's OEMed CD-ROMs, and Lexar memory key which Dell resells. Very likely not all transports or protocols are tested, for instance UFI. But I think it's right to call it "intensive" testing.
The main test is to put a CD and keyboard on a hub, and hub on a KVM, then flip KVM several times quickly from one blade to another. All hell breaks loose. IIRC, I had four different OOPS and lockup scenarios.
> Was there a reason to add more do-nothing code to host_reset?
Woopsie. I wanted to write it, but understood that if I return right code from "bus" reset, it should never be called. Sorry about that... I'll remove that part.
> Is it really safe to remove the irq_urb_sem?
The idea here is to have disconnect and resets locked against each other. They happen on different threads, unfortunately (khubd and scsi_eh). Initially I tried various orders, but then I thought, "Why am I making it hard on myself?! Much better just to merge them". The dev_semaphore now covers everything irq_urb_sem used to cover, except one path into usb_stor_allocate_irq from initial probing.
Thanks for looking at it!
-- Pete - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |