I thought I understood the way PCI devices are accessed in Linux,
but the code for the Promise PCI controllers has me confused. To
me it looks like the code is accessing registers outside the area
allocated to the PCI device during initialization.
The device is configured with the following resources (on a Mips-based
platform):
00:08.0 Class 0180: 105a:4d68 (rev 01)
I/O at 0x00000080 [size=0x8]
I/O at 0x00000088 [size=0x4]
I/O at 0x00000090 [size=0x8]
I/O at 0x00000098 [size=0x4]
I/O at 0x000000a0 [size=0x10]
Mem at 0x08004000 [size=0x4000]
The Promise code uses registers specified using bar4 but it uses offsets
above 0x0f. Here's an example of the code in drivers/ide/pdc202xx.c:
void pdc202xx_reset (ide_drive_t *drive)
{
unsigned long high_16 = pci_resource_start(HWIF(drive)->pci_dev, 4);
byte udma_speed_flag = inb(high_16 + 0x001f);
OUT_BYTE(udma_speed_flag | 0x10, high_16 + 0x001f);
mdelay(100);
OUT_BYTE(udma_speed_flag & ~0x10, high_16 + 0x001f);
mdelay(2000); /* 2 seconds ?! */
}
How can the code above try to write to a register 0x1f into a region with
a size of 0x10? Wouldn't that stomp on the registers of some other PCI device?
Thanks.
Gerald
I fixed that problem after it was discovered.
It is in the newest drivers.
On Wed, 16 Jan 2002, Gerald Champagne wrote:
> I thought I understood the way PCI devices are accessed in Linux,
> but the code for the Promise PCI controllers has me confused. To
> me it looks like the code is accessing registers outside the area
> allocated to the PCI device during initialization.
>
> The device is configured with the following resources (on a Mips-based
> platform):
>
> 00:08.0 Class 0180: 105a:4d68 (rev 01)
> I/O at 0x00000080 [size=0x8]
> I/O at 0x00000088 [size=0x4]
> I/O at 0x00000090 [size=0x8]
> I/O at 0x00000098 [size=0x4]
> I/O at 0x000000a0 [size=0x10]
> Mem at 0x08004000 [size=0x4000]
>
> The Promise code uses registers specified using bar4 but it uses offsets
> above 0x0f. Here's an example of the code in drivers/ide/pdc202xx.c:
>
> void pdc202xx_reset (ide_drive_t *drive)
> {
> unsigned long high_16 = pci_resource_start(HWIF(drive)->pci_dev, 4);
> byte udma_speed_flag = inb(high_16 + 0x001f);
>
> OUT_BYTE(udma_speed_flag | 0x10, high_16 + 0x001f);
> mdelay(100);
> OUT_BYTE(udma_speed_flag & ~0x10, high_16 + 0x001f);
> mdelay(2000); /* 2 seconds ?! */
> }
>
> How can the code above try to write to a register 0x1f into a region with
> a size of 0x10? Wouldn't that stomp on the registers of some other PCI device?
>
> Thanks.
>
> Gerald
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/
>
I just checked 2.5.3-pre1 and it still makes references to
registers above 0x0f. I see many references to registers
at 0x11, 0x1b, 0x1f. For example, the following:
unsigned long atapi_reg = high_16 + (hwif->channel ? 0x24 : 0x00);
That assumes atapi_reg will be 0x24 beyond the base of bar4 (high_16).
Is my PCI code showing me an incorrect size for bar4, or is this
accessing registers beyond its limits?
Thanks.
Gerald
Andre Hedrick wrote:
> I fixed that problem after it was discovered.
> It is in the newest drivers.
>
>
> On Wed, 16 Jan 2002, Gerald Champagne wrote:
>
>
>>I thought I understood the way PCI devices are accessed in Linux,
>>but the code for the Promise PCI controllers has me confused. To
>>me it looks like the code is accessing registers outside the area
>>allocated to the PCI device during initialization.
>>
>>The device is configured with the following resources (on a Mips-based
>>platform):
>>
>>00:08.0 Class 0180: 105a:4d68 (rev 01)
>> I/O at 0x00000080 [size=0x8]
>> I/O at 0x00000088 [size=0x4]
>> I/O at 0x00000090 [size=0x8]
>> I/O at 0x00000098 [size=0x4]
>> I/O at 0x000000a0 [size=0x10]
>> Mem at 0x08004000 [size=0x4000]
>>
>>The Promise code uses registers specified using bar4 but it uses offsets
>>above 0x0f. Here's an example of the code in drivers/ide/pdc202xx.c:
>>
>>void pdc202xx_reset (ide_drive_t *drive)
>>{
>> unsigned long high_16 = pci_resource_start(HWIF(drive)->pci_dev, 4);
>> byte udma_speed_flag = inb(high_16 + 0x001f);
>>
>> OUT_BYTE(udma_speed_flag | 0x10, high_16 + 0x001f);
>> mdelay(100);
>> OUT_BYTE(udma_speed_flag & ~0x10, high_16 + 0x001f);
>> mdelay(2000); /* 2 seconds ?! */
>>}
>>
>>How can the code above try to write to a register 0x1f into a region with
>>a size of 0x10? Wouldn't that stomp on the registers of some other PCI device?
>>
>>Thanks.
>>
>>Gerald
>>
>>-
>>To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
>>the body of a message to [email protected]
>>More majordomo info at http://vger.kernel.org/majordomo-info.html
>>Please read the FAQ at http://www.tux.org/lkml/
>>
>>
>
>
>
>
On Wed, 16 Jan 2002, Gerald Champagne wrote:
> I just checked 2.5.3-pre1 and it still makes references to
> registers above 0x0f. I see many references to registers
> at 0x11, 0x1b, 0x1f. For example, the following:
>
> unsigned long atapi_reg = high_16 + (hwif->channel ? 0x24 : 0x00);
>
> That assumes atapi_reg will be 0x24 beyond the base of bar4 (high_16).
> Is my PCI code showing me an incorrect size for bar4, or is this
> accessing registers beyond its limits?
Look in ide-pci.c
cat /proc/ioports | grep 20267
df00-df3f : Promise Technology, Inc. 20267
df10-df3f : PDC20267
dfa0-dfa7 : Promise Technology, Inc. 20267
dfa8-dfab : Promise Technology, Inc. 20267
dfac-dfaf : Promise Technology, Inc. 20267
dfe0-dfe7 : Promise Technology, Inc. 20267
pci bus 0x0 cardnum 0x03 function 0x0000: vendor 0x105a device 0x4d30
Device unknown
CardVendor 0x105a card 0x4d39 (Card unknown)
STATUS 0x0210 COMMAND 0x0007
CLASS 0x01 0x04 0x00 REVISION 0x02
BIST 0x00 HEADER 0x00 LATENCY 0x40 CACHE 0x00
BASE0 0x0000dfe1 addr 0x0000dfe0 I/O
BASE1 0x0000dfad addr 0x0000dfac I/O
BASE2 0x0000dfa1 addr 0x0000dfa0 I/O
BASE3 0x0000dfa9 addr 0x0000dfa8 I/O
BASE4 0x0000df01 addr 0x0000df00 I/O
BASE5 0xfeaa0000 addr 0xfeaa0000 MEM
BASEROM 0xfeae0001 addr 0xfeae0000 decode-enabled
MAX_LAT 0x00 MIN_GNT 0x00 INT_PIN 0x01 INT_LINE 0x0b
{DEVID_PDC20246,"PDC20246", PCI_PDC202XX, NULL,
INIT_PDC202XX, NULL, {{0x50,0x02,0x02}, {0x50,0x04,0x04}},
OFF_BOARD,
16 },
{DEVID_PDC20262,"PDC20262", PCI_PDC202XX, ATA66_PDC202XX,
INIT_PDC202XX, NULL, {{0x50,0x02,0x02}, {0x50,0x04,0x04}},
OFF_BOARD,
48 },
{DEVID_PDC20265,"PDC20265", PCI_PDC202XX, ATA66_PDC202XX,
INIT_PDC202XX, NULL, {{0x50,0x02,0x02}, {0x50,0x04,0x04}},
OFF_BOARD,
48 },
{DEVID_PDC20267,"PDC20267", PCI_PDC202XX, ATA66_PDC202XX,
INIT_PDC202XX, NULL, {{0x50,0x02,0x02}, {0x50,0x04,0x04}},
OFF_BOARD,
48 },
{DEVID_PDC20268,"PDC20268", PCI_PDC202XX, ATA66_PDC202XX,
INIT_PDC202XX, NULL, {{0x00,0x00,0x00}, {0x00,0x00,0x00}},
OFF_BOARD,
0 },
{DEVID_PDC20268R,"PDC20270", PCI_PDC202XX, ATA66_PDC202XX,
INIT_PDC202XX, NULL, {{0x00,0x00,0x00}, {0x00,0x00,0x00}},
OFF_BOARD,
0 },
{DEVID_PDC20269,"PDC20269", PCI_PDC202XX, ATA66_PDC202XX,
INIT_PDC202XX, NULL, {{0x00,0x00,0x00}, {0x00,0x00,0x00}},
OFF_BOARD,
0 },
{DEVID_PDC20275,"PDC20275", PCI_PDC202XX, ATA66_PDC202XX,
INIT_PDC202XX, NULL, {{0x00,0x00,0x00}, {0x00,0x00,0x00}},
OFF_BOARD,
0 },
int pdc202xx_dmaproc (ide_dma_action_t func, ide_drive_t *drive)
switch (dev->device) {
case PCI_DEVICE_ID_PROMISE_20275:
case PCI_DEVICE_ID_PROMISE_20269:
case PCI_DEVICE_ID_PROMISE_20268R:
case PCI_DEVICE_ID_PROMISE_20268:
newchip = 1;
break;
case PCI_DEVICE_ID_PROMISE_20267:
case PCI_DEVICE_ID_PROMISE_20265:
case PCI_DEVICE_ID_PROMISE_20262:
hardware48hack = 1;
clock = IN_BYTE(high_16 + 0x11);
default:
break;
}
See above that 62/65/67 have 48 extra io registers.
This it is legal to touch and use them.
Cheers,
Andre Hedrick
Linux Disk Certification Project Linux ATA Development