Jeff Garzik wrote:
Tejun Heo wrote:
The problem is that the timeout handler doesn't have anyway to
determine whether the timeout is from real timeout or from DMA error,
and the
Not true at all. Just read BMDMA status. Take a look at what
drivers/ide does.
Yeap, what I meant was the current timeout handler implementation
doesn't have any way to do that, so later in the previous reply, I
talked about ->timeout_autopsy.
timeout handler is responsible for transferring the ownership the
failed port to EH. EH, on entry, must be guaranteed that it owns the
port if it's not frozen.
One way around this would be making a new callback, say,
->timeout_autopsy and let it decide whether the port needs freezing or
not, but it would be an overkill. The only side effect of being
frozen is that the port will get a softreset to thaw it, which isn't
so bad - I want my controller to get a good spanking in the ass after
sitting idle for 30secs.
When presented with standard, documented DMA error behavior, a reset is
inappropriate. Just ACK the DMA error and move on with life. If
continuous DMA errors occur, reset and/or step down the speed as was
discussed many months ago.
The speeding down part is the same whether the port is frozen or not.
The only difference is how EH recovers the port after the error. Failed
devices on a not frozen port are just revalidated while a frozen port
gets a reset.
Here is another method to deal with it as adding ->timeout_autopsy or
anything similar is too unattractive. A new interface, say,
ata_eh_thaw_port() can be implemented which thaws the port without
resetting it. Then, in BMDMA autopsy, after determining that a timeout
was caused by DMA error, it can thaw the port and adjust qc->err_mask to
AC_ERR_HOST_BUS. How does it sound to you?
Get the user back up and talking to their disk as fast as possible.
Command timeout is 30 secs (which, I think is a bit too long for ATA
disk devices). If resetting succeeds, it takes less than two seconds.
I don't think it will make any difference to the user.
--
tejun
-
: send the line "unsubscribe linux-ide" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
