Hi, Reza
IMHO, it should jump out the loop at the below point if we put break statement
inside the new if.
PrdCount = *(volatile UINT32 *)
(&(AhciRegisters->AhciCmdList[0].AhciCmdPrdbc));
if (PrdCount == DataCount) {
break;
}
As I have no the test env, could you help me verify if it's true?
Thanks
Feng
-----Original Message-----
From: Reza Jelveh [mailto:[email protected]]
Sent: Friday, August 08, 2014 17:54
To: Tian, Feng
Cc: [email protected]; [email protected]
Subject: Re: [edk2] [PATCH 2/2] MdeModulePkg: Check D2H register status in
AhciPioTransfer
On 08/08/14 01:12, Tian, Feng wrote:
> Hi, Reza
>
> That's the problem.
>
> For PIO DATA-OUT command protocol, D2H either indicates that the transfer was
> completed or that's an error occurred.
> But for PIO DATA-IN command protocol, I don't see such sentence from SATA 2.6
> spec.
>
> If the transfer has been completed, why the loop doesn't jump out when SETUP
> FIS received and expected data size was transferred?
> So there is only one possibility: more data transfer required but D2H
> received here. How could we return now? Doesn't it cause data lost?
If we don't break out of the loop the opposite will happen. It will continue
pausing and reading indefinitely until the retry limit is reached.
I'm not sure if I'm right but if you were to follow the spec literally
responses to commands like identify are synchronous and don't have a D2H at the
end. OS drivers seem to have no problems with these quirks, but AhciMode.c will
simply fail with those sata controllers.
I'm guessing that some Arm controller have the same problem after seeing this
patch on the linaro mailing list:
http://lists.linaro.org/pipermail/linaro-uefi/2014-June/000275.html
According to the spec D2H here should indeed only occur on error, but this is
not always the case.
------------------------------------------------------------------------------
_______________________________________________
edk2-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/edk2-devel
