David and Alan, The patch below looks sane to me, do you mind if I clean it up to match the proper coding guidelines and add it to the tree? Saito, thank you very much for doing this work, can you provide a "Signed-off-by:" line, as documented in the file, Documentation/SubmittingPatches so that I can apply this patch to the kernel tree? thanks, greg k-h On Thu, Jan 08, 2009 at 06:37:25PM +0900, Saito.Koichiro@xxxxxxxxx wrote: > I have the following problem. My machine's chipset is > ICH5. I plugged in a USB flash drive and a DVD drive. I > start to access USB flash drive by a CP command and while > it's running, unplugged a dvd drive. CP command running is > failed. > > I got advice about this problem from Alan-san, Matthew-san > and David-san on anther thread. > http://markmail.org/message/k2drtwwkul6dawp2#query:koichiro%20saito%20usb%20linux+page:1+mid:wn27qdeqbqwgoaes+state:results > I obtained information that this problem occur by Alain's > system (Intel 82845GE motherboard with ICH4 chipset). The > cause of this error is a common hardware bug among EHCI > controllers. This error occur by XactErr. > > I examined how Windows work around with this problem. > Windows corrected with a following HOTFIX. But the content > of the symptom is different. > http://support.microsoft.com/KB/908673 > When I install this HOTFIX in my windows machine, I > confirmed this problem is resolved. The description of > HOTFIX said that When XactErr occurred, ehci host driver of > Windows run soft retrying. > > From the aforementioned process, I try to modified > ehci-driver to retry when XactErr occurred. > Could you give me any advice whether this soft retrying is > valid and correct? > > > diff -urN linux-2.6.28-git10.org/drivers/usb/host/ehci-q.c linux-2.6.28-git10/drivers/usb/host/ehci-q.c > --- linux-2.6.28-git10.org/drivers/usb/host/ehci-q.c 2008-12-25 08:26:37.000000000 +0900 > +++ linux-2.6.28-git10/drivers/usb/host/ehci-q.c 2009-01-09 06:46:05.000000000 +0900 > @@ -341,6 +341,35 @@ > if ((token & QTD_STS_HALT) != 0) { > stopped = 1; > > + /* judge transaction error > + * "three strikes and you're out" policy > + * is applied to retrying */ > + if (token & QTD_STS_XACT && > + !QTD_CERR (token) && > + --qh->xacterrs > 0) { > + ehci_dbg (ehci, "@@@ detected XactErr len %d/%d retry %d\n", > + qtd->length - QTD_LENGTH(token), qtd->length, > + QH_XACTERR_MAX - qh->xacterrs); > + /* stop flag off*/ > + stopped = 0; > + > + /* token clear error status for tranxaction error retry*/ > + /* clear token status */ > + token &= 0xffffff00; /* status -> off */ > + token |= QTD_STS_ACTIVE; /* Active -> on */ > + token |= (EHCI_TUNE_CERR << 10); /* clear CERR */ > + /* update qtd token */ > + qtd->hw_token = cpu_to_hc32(ehci, token); > + wmb (); > + > + /* update qh */ > + qh->qh_state = QH_STATE_IDLE; > + qh_update(ehci, qh, qtd); > + > + break; > + } > + > + > /* magic dummy for some short reads; qh won't advance. > * that silicon quirk can kick in with this dummy too. > * > @@ -949,6 +978,8 @@ > > urb->hcpriv = qh_get (qh); > } > + /* set transaction error retry counter */ > + qh->xacterrs = QH_XACTERR_MAX; > } > return qh; > } > diff -urN linux-2.6.28-git10.org/drivers/usb/host/ehci.h linux-2.6.28-git10/drivers/usb/host/ehci.h > --- linux-2.6.28-git10.org/drivers/usb/host/ehci.h 2009-01-09 05:55:34.000000000 +0900 > +++ linux-2.6.28-git10/drivers/usb/host/ehci.h 2009-01-09 06:46:05.000000000 +0900 > @@ -379,6 +379,13 @@ > unsigned short start; /* where polling starts */ > #define NO_FRAME ((unsigned short)~0) /* pick new start */ > struct usb_device *dev; /* access to TT */ > + u8 xacterrs; /* XactErr retry counter */ > +#define QH_XACTERR_MAX 3 /* XactErr retry limit > + * "three strikes and > + * you're out" policy > + * is applied to retrying > + */ > + > } __attribute__ ((aligned (32))); > > /*-------------------------------------------------------------------------*/ > > > best regards > Koichiro Saito -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
