| Message ID | 1459956836-8213-1-git-send-email-chunkeey@googlemail.com |
|---|---|
| State | Not Applicable |
| Delegated to: | David Miller |
| Headers | show |
On Wed, Apr 6, 2016 at 6:33 PM, Christian Lamparter <chunkeey@googlemail.com> wrote: > This patch allows sata_dwc_460ex to work with non-ncq, dma i/o. > Previously, the driver dumped the following warning: > > ------------[ cut here ]------------ > WARNING: at c01f3d50 [verbose debug info unavailable] > CPU: 0 PID: 315 Comm: kworker/u2:2 Not tainted 4.4.6 #18 > Workqueue: events_unbound async_run_entry_fn > NIP [c01f3d50] ata_sff_qc_issue+0x1d8/0x204 > LR [c01f3bbc] ata_sff_qc_issue+0x44/0x204 > Call Trace: > [cf6a98c0] [c01f3bbc] ata_sff_qc_issue+0x44/0x204 (unreliable) > [cf6a98e0] [c01f4e38] sata_dwc_qc_issue+0x270/0x298 > [cf6a9950] [c01e24f4] ata_qc_issue+0x2f0/0x364 > > and was unable to access the drive. The issue is that > ata_sff_qc_issue can only handle PIO. This patch replaces > the it with ata_bmdma_qc_issue which handles DMA transfers > and can fallback to PIO when needed. > > Cc: stable@vger.kernel.org > Signed-off-by: Christian Lamparter <chunkeey@googlemail.com> > --- > Note: I have a system with a APM82181 (which is a ppc460ex derivative). > It's the SoC in the WD MyBook Live NAS. > > The driver has other issues as well... e.g: > > Unable to handle kernel paging request for data at address 0x00000004 > Faulting instruction address: 0xc01f2640 > Oops: Kernel access of bad area, sig: 11 [#1] > CPU: 0 PID: 1 Comm: swapper Not tainted 4.4.6 #42 > NIP [c01f2640] sata_dwc_probe+0x1a8/0x2dc > LR [c01f2640] sata_dwc_probe+0x1a8/0x2dc > Call Trace: > [cf435d60] [c01f2640] sata_dwc_probe+0x1a8/0x2dc (unreliable) > [cf435dc0] [c01c5728] platform_drv_probe+0x34/0x70 > (This is due to hsdev->dma being NULL, as it was never allocated > or set). > > However, this has been fixed (including fixes to dwc's > DMA) by Mans Rullgard. His repo is available from: > <https://bitbucket.org/mansr/linux-dwc/branch/dwc-sata> > > Note2: NCQ seems to freezes the controller. Not sure what's > going on. I can post a separate patch to disable it, if requested. Thanks for the patch. Like I said before please check the branch we have and jump in in testing / suggesting better fixes. Please, also keep Måns and me in Cc list regarding this driver. > --- > drivers/ata/sata_dwc_460ex.c | 4 +++- > 1 file changed, 3 insertions(+), 1 deletion(-) > > diff --git a/drivers/ata/sata_dwc_460ex.c b/drivers/ata/sata_dwc_460ex.c > index e7a47aa..234f50d 100644 > --- a/drivers/ata/sata_dwc_460ex.c > +++ b/drivers/ata/sata_dwc_460ex.c > @@ -1113,7 +1113,9 @@ static unsigned int sata_dwc_qc_issue(struct ata_queued_cmd *qc) > sata_dwc_exec_command_by_tag(ap, &qc->tf, qc->tag, > SATA_DWC_CMD_ISSUED_PEND); > } else { > - ata_sff_qc_issue(qc); > + /* Sync ata_port with qc->tag */ > + ap->link.active_tag = qc->tag; > + return ata_bmdma_qc_issue(qc); > } > return 0; > } > -- > 2.8.0.rc3 >
diff --git a/drivers/ata/sata_dwc_460ex.c b/drivers/ata/sata_dwc_460ex.c index e7a47aa..234f50d 100644 --- a/drivers/ata/sata_dwc_460ex.c +++ b/drivers/ata/sata_dwc_460ex.c @@ -1113,7 +1113,9 @@ static unsigned int sata_dwc_qc_issue(struct ata_queued_cmd *qc) sata_dwc_exec_command_by_tag(ap, &qc->tf, qc->tag, SATA_DWC_CMD_ISSUED_PEND); } else { - ata_sff_qc_issue(qc); + /* Sync ata_port with qc->tag */ + ap->link.active_tag = qc->tag; + return ata_bmdma_qc_issue(qc); } return 0; }
This patch allows sata_dwc_460ex to work with non-ncq, dma i/o. Previously, the driver dumped the following warning: ------------[ cut here ]------------ WARNING: at c01f3d50 [verbose debug info unavailable] CPU: 0 PID: 315 Comm: kworker/u2:2 Not tainted 4.4.6 #18 Workqueue: events_unbound async_run_entry_fn NIP [c01f3d50] ata_sff_qc_issue+0x1d8/0x204 LR [c01f3bbc] ata_sff_qc_issue+0x44/0x204 Call Trace: [cf6a98c0] [c01f3bbc] ata_sff_qc_issue+0x44/0x204 (unreliable) [cf6a98e0] [c01f4e38] sata_dwc_qc_issue+0x270/0x298 [cf6a9950] [c01e24f4] ata_qc_issue+0x2f0/0x364 and was unable to access the drive. The issue is that ata_sff_qc_issue can only handle PIO. This patch replaces the it with ata_bmdma_qc_issue which handles DMA transfers and can fallback to PIO when needed. Cc: stable@vger.kernel.org Signed-off-by: Christian Lamparter <chunkeey@googlemail.com> --- Note: I have a system with a APM82181 (which is a ppc460ex derivative). It's the SoC in the WD MyBook Live NAS. The driver has other issues as well... e.g: Unable to handle kernel paging request for data at address 0x00000004 Faulting instruction address: 0xc01f2640 Oops: Kernel access of bad area, sig: 11 [#1] CPU: 0 PID: 1 Comm: swapper Not tainted 4.4.6 #42 NIP [c01f2640] sata_dwc_probe+0x1a8/0x2dc LR [c01f2640] sata_dwc_probe+0x1a8/0x2dc Call Trace: [cf435d60] [c01f2640] sata_dwc_probe+0x1a8/0x2dc (unreliable) [cf435dc0] [c01c5728] platform_drv_probe+0x34/0x70 (This is due to hsdev->dma being NULL, as it was never allocated or set). However, this has been fixed (including fixes to dwc's DMA) by Mans Rullgard. His repo is available from: <https://bitbucket.org/mansr/linux-dwc/branch/dwc-sata> Note2: NCQ seems to freezes the controller. Not sure what's going on. I can post a separate patch to disable it, if requested. --- drivers/ata/sata_dwc_460ex.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)