2003-09-18 04:00:27

by Allen Martin

[permalink] [raw]
Subject: [PATCH] 2.4.23-pre4 add support for udma6 to nForce IDE driver

This adds support for udma6 (Ultra 133) to the combined AMD / NVIDIA IDE
driver. It depends on linux-2.4.23-pre4-nvide.patch.

It looks like UDMA modes > udma2 were shoehorned into this driver, so the
code to set the UDMA timing registers is pretty ugly. In particular,
computing UDMA cycle times from UDMA mode and then mapping that into a
register value is akward and cause for much of the complexity in
amd_set_speed(). At least on all the nForce controllers, this register is
always programmed the same, regardless if the PCI bus is 33MHz or 66MHz, so
the UDMA cycle times can be ignored.



diff -ru -X dontdiff linux-2.4.23-pre4-nvide/drivers/ide/pci/amd74xx.c
linux-2.4.23-pre4-nvata133/drivers/ide/pci/amd74xx.c
--- linux-2.4.23-pre4-nvide/drivers/ide/pci/amd74xx.c 2003-09-17
19:58:46.000000000 -0700
+++ linux-2.4.23-pre4-nvata133/drivers/ide/pci/amd74xx.c 2003-09-17
20:11:20.000000000 -0700
@@ -40,6 +40,7 @@
#define AMD_UDMA_33 0x01
#define AMD_UDMA_66 0x02
#define AMD_UDMA_100 0x03
+#define AMD_UDMA_133 0x04
#define AMD_CHECK_SWDMA 0x08
#define AMD_BAD_SWDMA 0x10
#define AMD_BAD_FIFO 0x20
@@ -60,13 +61,13 @@
{ PCI_DEVICE_ID_AMD_OPUS_7441, 0x00, 0x40, AMD_UDMA_100 },
/* AMD-768 Opus */
{ PCI_DEVICE_ID_AMD_8111_IDE, 0x00, 0x40, AMD_UDMA_100 },
/* AMD-8111 */
{ PCI_DEVICE_ID_NVIDIA_NFORCE_IDE, 0x00, 0x50, AMD_UDMA_100 },
/* nVidia nForce */
- { PCI_DEVICE_ID_NVIDIA_NFORCE2_IDE, 0x00, 0x50, AMD_UDMA_100 },
/* nVidia nForce2 */
- { PCI_DEVICE_ID_NVIDIA_NFORCE2S_IDE, 0x00, 0x50, AMD_UDMA_100 },
/* nVidia nForce2s */
- { PCI_DEVICE_ID_NVIDIA_NFORCE2S_SATA, 0x00, 0x50, AMD_UDMA_100 },
/* nVidia nForce2s SATA */
- { PCI_DEVICE_ID_NVIDIA_NFORCE3_IDE, 0x00, 0x50, AMD_UDMA_100 },
/* NVIDIA nForce3 */
- { PCI_DEVICE_ID_NVIDIA_NFORCE3S_IDE, 0x00, 0x50, AMD_UDMA_100 },
/* NVIDIA nForce3s */
- { PCI_DEVICE_ID_NVIDIA_NFORCE3S_SATA, 0x00, 0x50, AMD_UDMA_100 },
/* NVIDIA nForce3s SATA */
- { PCI_DEVICE_ID_NVIDIA_NFORCE3S_SATA2, 0x00, 0x50, AMD_UDMA_100 },
/* NVIDIA nForce3s SATA2 */
+ { PCI_DEVICE_ID_NVIDIA_NFORCE2_IDE, 0x00, 0x50, AMD_UDMA_133 },
/* nVidia nForce2 */
+ { PCI_DEVICE_ID_NVIDIA_NFORCE2S_IDE, 0x00, 0x50, AMD_UDMA_133 },
/* nVidia nForce2s */
+ { PCI_DEVICE_ID_NVIDIA_NFORCE2S_SATA, 0x00, 0x50, AMD_UDMA_133 },
/* nVidia nForce2s SATA */
+ { PCI_DEVICE_ID_NVIDIA_NFORCE3_IDE, 0x00, 0x50, AMD_UDMA_133 },
/* NVIDIA nForce3 */
+ { PCI_DEVICE_ID_NVIDIA_NFORCE3S_IDE, 0x00, 0x50, AMD_UDMA_133 },
/* NVIDIA nForce3s */
+ { PCI_DEVICE_ID_NVIDIA_NFORCE3S_SATA, 0x00, 0x50, AMD_UDMA_133 },
/* NVIDIA nForce3s SATA */
+ { PCI_DEVICE_ID_NVIDIA_NFORCE3S_SATA2, 0x00, 0x50, AMD_UDMA_133 },
/* NVIDIA nForce3s SATA2 */

{ 0 }
};
@@ -76,9 +77,9 @@
static unsigned int amd_80w;
static unsigned int amd_clock;

-static unsigned char amd_cyc2udma[] = { 6, 6, 5, 4, 0, 1, 1, 2, 2, 3, 3 };
-static unsigned char amd_udma2cyc[] = { 4, 6, 8, 10, 3, 2, 1, 1 };
-static char *amd_dma[] = { "MWDMA16", "UDMA33", "UDMA66", "UDMA100" };
+static unsigned char amd_cyc2udma[] = { 6, 6, 5, 4, 0, 1, 1, 2, 2, 3, 3, 7
};
+static unsigned char amd_udma2cyc[] = { 4, 6, 8, 10, 3, 2, 1, 0 };
+static char *amd_dma[] = { "MWDMA16", "UDMA33", "UDMA66", "UDMA100",
"UDMA133" };

/*
* AMD /proc entry.
@@ -160,6 +161,11 @@
cycle[i] = 666666 / amd_clock;
continue;
}
+ if (den[i] && uen[i] && udma[i] == 0) {
+ speed[i] = amd_clock * 4;
+ cycle[i] = 500000 / amd_clock;
+ continue;
+ }

speed[i] = 4 * amd_clock / ((den[i] && uen[i]) ? udma[i] :
(active[i] + recover[i]) * 2);
cycle[i] = 1000000 * ((den[i] && uen[i]) ? udma[i] :
(active[i] + recover[i]) * 2) / amd_clock / 2;
@@ -206,6 +212,7 @@
case AMD_UDMA_33: t = timing->udma ? (0xc0 |
(FIT(timing->udma, 2, 5) - 2)) : 0x03; break;
case AMD_UDMA_66: t = timing->udma ? (0xc0 |
amd_cyc2udma[FIT(timing->udma, 2, 10)]) : 0x03; break;
case AMD_UDMA_100: t = timing->udma ? (0xc0 |
amd_cyc2udma[FIT(timing->udma, 1, 10)]) : 0x03; break;
+ case AMD_UDMA_133: t = timing->udma ? (0xc0 |
amd_cyc2udma[FIT(timing->udma, 1, 11)]) : 0x03; break;
default: return;
}

@@ -239,7 +246,12 @@
ide_timing_merge(&p, &t, &t, IDE_TIMING_8BIT);
}

+ /*
+ * AMD / nForce UDMA timing register should really be programmed
+ * based on UDMA mode not UDMA cycle time...
+ */
if (speed == XFER_UDMA_5 && amd_clock <= 33333) t.udma = 1;
+ if (speed == XFER_UDMA_6 && amd_clock <= 33333) t.udma = 11;

amd_set_speed(HWIF(drive)->pci_dev, drive->dn, &t);

@@ -282,7 +294,8 @@
XFER_PIO | XFER_EPIO | XFER_MWDMA | XFER_UDMA |
((amd_config->flags & AMD_BAD_SWDMA) ? 0 : XFER_SWDMA) |
(w80 && (amd_config->flags & AMD_UDMA) >= AMD_UDMA_66 ?
XFER_UDMA_66 : 0) |
- (w80 && (amd_config->flags & AMD_UDMA) >= AMD_UDMA_100 ?
XFER_UDMA_100 : 0));
+ (w80 && (amd_config->flags & AMD_UDMA) >= AMD_UDMA_100 ?
XFER_UDMA_100 : 0) |
+ (w80 && (amd_config->flags & AMD_UDMA) >= AMD_UDMA_133 ?
XFER_UDMA_133 : 0));

amd_set_drive(drive, speed);

@@ -318,6 +331,7 @@

switch (amd_config->flags & AMD_UDMA) {

+ case AMD_UDMA_133:
case AMD_UDMA_100:
pci_read_config_byte(dev, AMD_CABLE_DETECT, &t);
pci_read_config_dword(dev, AMD_UDMA_TIMING, &u);


2003-09-19 22:58:12

by Merlin Hughes

[permalink] [raw]
Subject: RE: [PATCH] 2.4.23-pre4 add support for udma6 to nForce IDE driver


Hi Allen,

Do you know if an expected side-effect of this patch would
be increased stability? I see that the code does more than
just adding in a new UDMA number, I'm just wondering if the
other changes could be the deliberate cause of the reliability
I'm now seeing.

I have an nforce2 motherboard (Shuttle) with a Maxtor
UDMA133 drive. Operations such as dd if=/dev/hda of=/dev/null
would reliably freeze the machine under 2.4.23-pre4 at
UDMA100; I had to drop to UDMA44 to overcome this.

Since applying your patch, however, I've managed to run such
a dd, with zcav thrown in, with complete relability at UDMA133
for several hours without problems.

Thanks, Merlin