Message ID | CAO1A-8A0c=5GnH4wLYBz+N1tHLUO3CeuOSq_7S+ZxscNqt5Hnw@mail.gmail.com |
---|---|
State | Not Applicable |
Delegated to: | David Miller |
Headers | show |
On Fri, Apr 12, 2013 at 12:34:24PM -0700, Marc C wrote: > Hello Jeff, .. > index 4e73e1b..383de21 100644 > --- a/linux/drivers/ata/libata-core.c > +++ b/linux/drivers/ata/libata-core.c > @@ -4827,6 +4827,12 @@ void ata_qc_complete(struct ata_queued_cmd *qc) > { > struct ata_port *ap = qc->ap; > > + if (ata_is_atapi(qc->tf.protocol)) { > + void ahci_apply_atapi_workaround(struct ata_queued_cmd *qc); This prototype really doesn't belong here. Should probably be in include/linux/libata.h
On 04/12/2013 01:34 PM, Marc C wrote: > Hello Jeff, > > I'm working with a proprietary but AHCI-compatible SATA controller that uses > the libata driver. Early versions of this controller have issues with processing > non-data ATAPI commands. A workaround was identified which requires some > register pokes in the command completion path of the driver. I don't expect to > push this patch upstream (yet). However, I would like to get some feedback > regarding the workaround, and check if the placement of the code is "acceptable" > or if there would be a better place to put it. > > The workaround itself is rather simple: toggle the START bit immediately after a > non-data ATAPI command completes. The ST bit toggle would be performed within > atomic context in the AHCI port interrupt ISR. > > Signed-off-by: Marc C <marc.ceeeee@gmail.com> This should likely be triggered off a host flag like some of the ones already in the driver so it can be done just for the affected controllers, not unconditionally. > --- > linux/drivers/ata/libahci.c | 26 ++++++++++++++++++++++++++ > linux/drivers/ata/libata-core.c | 6 ++++++ > 2 files changed, 32 insertions(+) > > diff --git a/linux/drivers/ata/libahci.c b/linux/drivers/ata/libahci.c > index f9eaa82..20b3a9a 100644 > --- a/linux/drivers/ata/libahci.c > +++ b/linux/drivers/ata/libahci.c > @@ -2215,6 +2215,32 @@ void ahci_set_em_messages(struct ahci_host_priv *hpriv, > } > EXPORT_SYMBOL_GPL(ahci_set_em_messages); > > +void ahci_apply_atapi_workaround(struct ata_queued_cmd *qc) > +{ > + if (!(qc->flags & ATA_QCFLAG_DMAMAP)) { > + /* > + * Non-data ATAPI commands need to kick the DMA engine > + */ > + struct ata_port *ap = qc->ap; > + void __iomem *port_mmio = ahci_port_base(ap); > + u32 tmp = readl(port_mmio + PORT_CMD); > + > + /* check if the HBA is idle */ > + if ((tmp & (PORT_CMD_START | PORT_CMD_LIST_ON)) != 0) { > + /* setting HBA to idle */ > + tmp &= ~PORT_CMD_START; > + writel(tmp, port_mmio + PORT_CMD); > + > + /* wait for engine to stop */ > + do { > + tmp = readl(port_mmio + PORT_CMD); > + } while (tmp & PORT_CMD_LIST_ON); > + } > + ahci_start_engine(ap); > + } > +} > +EXPORT_SYMBOL_GPL(ahci_apply_atapi_workaround); > + > MODULE_AUTHOR("Jeff Garzik"); > MODULE_DESCRIPTION("Common AHCI SATA low-level routines"); > MODULE_LICENSE("GPL"); > diff --git a/linux/drivers/ata/libata-core.c b/linux/drivers/ata/libata-core.c > index 4e73e1b..383de21 100644 > --- a/linux/drivers/ata/libata-core.c > +++ b/linux/drivers/ata/libata-core.c > @@ -4827,6 +4827,12 @@ void ata_qc_complete(struct ata_queued_cmd *qc) > { > struct ata_port *ap = qc->ap; > > + if (ata_is_atapi(qc->tf.protocol)) { > + void ahci_apply_atapi_workaround(struct ata_queued_cmd *qc); > + ahci_apply_atapi_workaround(qc); > + } > + > /* XXX: New EH and old EH use different mechanisms to > * synchronize EH with regular execution path. > * > -- > 1.7.9.5 > > > > > -- > Regards, > Marc > marc.ceeeee@gmail.com > -- > To unsubscribe from this list: send the line "unsubscribe linux-ide" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html > -- To unsubscribe from this list: send the line "unsubscribe linux-ide" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 04/13/2013 01:05 AM, Robert Hancock wrote: > On 04/12/2013 01:34 PM, Marc C wrote: >> Hello Jeff, >> >> I'm working with a proprietary but AHCI-compatible SATA controller >> that uses >> the libata driver. Early versions of this controller have issues with >> processing >> non-data ATAPI commands. A workaround was identified which requires some >> register pokes in the command completion path of the driver. I don't >> expect to >> push this patch upstream (yet). However, I would like to get some >> feedback >> regarding the workaround, and check if the placement of the code is >> "acceptable" >> or if there would be a better place to put it. >> >> The workaround itself is rather simple: toggle the START bit >> immediately after a >> non-data ATAPI command completes. The ST bit toggle would be performed >> within >> atomic context in the AHCI port interrupt ISR. >> >> Signed-off-by: Marc C <marc.ceeeee@gmail.com> > > This should likely be triggered off a host flag like some of the ones > already in the driver so it can be done just for the affected > controllers, not unconditionally. Indeed -- you don't want to punish the non-buggy 99% case for one weird buggy hardware rev. See AHCI_HFLAG_xxx Jeff -- To unsubscribe from this list: send the line "unsubscribe linux-ide" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/linux/drivers/ata/libahci.c b/linux/drivers/ata/libahci.c index f9eaa82..20b3a9a 100644 --- a/linux/drivers/ata/libahci.c +++ b/linux/drivers/ata/libahci.c @@ -2215,6 +2215,32 @@ void ahci_set_em_messages(struct ahci_host_priv *hpriv, } EXPORT_SYMBOL_GPL(ahci_set_em_messages); +void ahci_apply_atapi_workaround(struct ata_queued_cmd *qc) +{ + if (!(qc->flags & ATA_QCFLAG_DMAMAP)) { + /* + * Non-data ATAPI commands need to kick the DMA engine + */ + struct ata_port *ap = qc->ap; + void __iomem *port_mmio = ahci_port_base(ap); + u32 tmp = readl(port_mmio + PORT_CMD); + + /* check if the HBA is idle */ + if ((tmp & (PORT_CMD_START | PORT_CMD_LIST_ON)) != 0) { + /* setting HBA to idle */ + tmp &= ~PORT_CMD_START; + writel(tmp, port_mmio + PORT_CMD); + + /* wait for engine to stop */ + do { + tmp = readl(port_mmio + PORT_CMD); + } while (tmp & PORT_CMD_LIST_ON); + } + ahci_start_engine(ap); + } +} +EXPORT_SYMBOL_GPL(ahci_apply_atapi_workaround); + MODULE_AUTHOR("Jeff Garzik"); MODULE_DESCRIPTION("Common AHCI SATA low-level routines"); MODULE_LICENSE("GPL"); diff --git a/linux/drivers/ata/libata-core.c b/linux/drivers/ata/libata-core.c index 4e73e1b..383de21 100644 --- a/linux/drivers/ata/libata-core.c +++ b/linux/drivers/ata/libata-core.c @@ -4827,6 +4827,12 @@ void ata_qc_complete(struct ata_queued_cmd *qc) { struct ata_port *ap = qc->ap; + if (ata_is_atapi(qc->tf.protocol)) { + void ahci_apply_atapi_workaround(struct ata_queued_cmd *qc); + ahci_apply_atapi_workaround(qc); + } + /* XXX: New EH and old EH use different mechanisms to * synchronize EH with regular execution path. * -- 1.7.9.5
Hello Jeff, I'm working with a proprietary but AHCI-compatible SATA controller that uses the libata driver. Early versions of this controller have issues with processing non-data ATAPI commands. A workaround was identified which requires some register pokes in the command completion path of the driver. I don't expect to push this patch upstream (yet). However, I would like to get some feedback regarding the workaround, and check if the placement of the code is "acceptable" or if there would be a better place to put it. The workaround itself is rather simple: toggle the START bit immediately after a non-data ATAPI command completes. The ST bit toggle would be performed within atomic context in the AHCI port interrupt ISR. Signed-off-by: Marc C <marc.ceeeee@gmail.com> --- linux/drivers/ata/libahci.c | 26 ++++++++++++++++++++++++++ linux/drivers/ata/libata-core.c | 6 ++++++ 2 files changed, 32 insertions(+) -- Regards, Marc marc.ceeeee@gmail.com -- To unsubscribe from this list: send the line "unsubscribe linux-ide" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html