DMA appears broken in 2.6.11 for Mac 7200

DMA appears broken in 2.6.11 for Mac 7200

[linuxppcdev]

Just updated a couple of old Mac 7200 (601) boxen and ran into trouble with
the ide driver and a promise PDC20267 card. 

The very same drives and card work fine on a 9500 but hang on a 7200. If
I set the kernel arg ide=nodma I get in the dmesg:
ide_setup: ide=nodma : Prevented DMA
and the 7200 boots and runs fine.

Another suspect, the mace ethernet driver (which also appears to use DMA)
doesn't work on the 7200 with 2.6.11 but does with 2.4.28 and it works 
with 2.6.11 on the 9500.

Finally, I don't think this is related to DMA however the OF frame buffer
driver, if compiled into the kernel, kills the platinumfb driver which wasn't
a problem in the 2.4 kernels either (I get a blank screen but the machine
keeps running).

Bob

Re: DMA appears broken in 2.6.11 for Mac 7200

[Benjamin Herrenschmidt]

On Mon, 2005-03-14 at 03:51 +0000, linuxppcdev@qbjnet.com wrote:
> Just updated a couple of old Mac 7200 (601) boxen and ran into trouble with
> the ide driver and a promise PDC20267 card. 
> 
> The very same drives and card work fine on a 9500 but hang on a 7200. If
> I set the kernel arg ide=nodma I get in the dmesg:
> ide_setup: ide=nodma : Prevented DMA
> and the 7200 boots and runs fine.
> 
> Another suspect, the mace ethernet driver (which also appears to use DMA)
> doesn't work on the 7200 with 2.6.11 but does with 2.4.28 and it works 
> with 2.6.11 on the 9500.

Interesting. Not sure what's up, though. Those old machines were known
to have bugs relative to cache coherency... also check wetehr we are
setting the cache line size properly in PCI devices. You can also try to
disable use of PCI memory write & invalidate command in all devices.
 
> Finally, I don't think this is related to DMA however the OF frame buffer
> driver, if compiled into the kernel, kills the platinumfb driver which wasn't
> a problem in the 2.4 kernels either (I get a blank screen but the machine
> keeps running).

Ah ? platinumfb alone works, and platinumfb + offb fails ? weird ...

Ben.

Re: DMA appears broken in 2.6.11 for Mac 7200

[linuxppcdev]

Ben Said:
> On Mon, 2005-03-14 at 03:51 +0000, linuxppcdev@qbjnet.com wrote:
> > Just updated a couple of old Mac 7200 (601) boxen and ran into trouble with
> > the ide driver and a promise PDC20267 card. 
> > 
> > The very same drives and card work fine on a 9500 but hang on a 7200. If
> > I set the kernel arg ide=nodma I get in the dmesg:
> > ide_setup: ide=nodma : Prevented DMA
> > and the 7200 boots and runs fine.
> > 
> > Another suspect, the mace ethernet driver (which also appears to use DMA)
> > doesn't work on the 7200 with 2.6.11 but does with 2.4.28 and it works 
> > with 2.6.11 on the 9500.
> 
> Interesting. Not sure what's up, though. Those old machines were known
> to have bugs relative to cache coherency... also check wetehr we are
> setting the cache line size properly in PCI devices. You can also try to
> disable use of PCI memory write & invalidate command in all devices.
 
Here is some info from a 7200 booted off of 2.6.11:

dmesg

Total memory = 112MB; using 256kB for hash table (at c04c0000)
Linux version 2.6.11.2 (root@dev) (gcc version 3.3.4 (Debian 1:3.3.4-9ubuntu5)) 
#8 Sun Mar 13 12:50:26 CST 2005
Found a Grand Central mac-io controller, rev: 2, mapped at 0xfdf00000
PowerMac motherboard: PowerMac 7200/7300
Cache coherency enabled for bandit/PSX
Found Bandit PCI host bridge at 0xf2000000. Firmware bus number: 0->0
nvram: OF partition at 0x1800
nvram: XP partition at 0x1300
nvram: NR partition at 0x1400
On node 0 totalpages: 28672
  DMA zone: 28672 pages, LIFO batch:7
  Normal zone: 0 pages, LIFO batch:1
  HighMem zone: 0 pages, LIFO batch:1
Built 1 zonelists
Kernel command line:  root=/dev/hda2 video=platinumfb ide=nodma
ide_setup: ide=nodma : Prevented DMA
System has 32 possible interrupts
PID hash table entries: 512 (order: 9, 8192 bytes)
.
.
.
Forcing PCI IDE into native mode: 0000:00:0d.0
Rewrite of PROGIF failed !
.
.
PDC20267: IDE controller at PCI slot 0000:00:0d.0
PDC20267: chipset revision 2
PDC20267: ROM enabled at 0x80860000
PDC20267: 100% native mode on irq 23
PDC20267: (U)DMA Burst Bit DISABLED Primary PCI Mode Secondary PCI Mode.
    ide0: BM-DMA at 0x0480-0x0487, BIOS settings: hda:pio, hdb:pio
    ide1: BM-DMA at 0x0488-0x048f, BIOS settings: hdc:pio, hdd:pio
Probing IDE interface ide0...
.
.
eth0: Digital DS21143 Tulip rev 33 at c805a000, 00:40:F6:84:F8:C0, IRQ 25.
pmac: can't request resource 0!

lspci -vv

0000:00:0b.0 Host bridge: Apple Computer Inc. Bandit PowerPC host bridge (rev 03)
        Control: I/O- Mem+ BusMaster+ SpecCycle- MemWINV+ VGASnoop- ParErr- Stepping- SERR- FastB2B-
        Status: Cap- 66MHz- UDF- FastB2B+ ParErr- DEVSEL=medium >TAbort- <TAbort- <MAbort+ >SERR- <PERR-
        Latency: 32, Cache Line Size: 0x08 (32 bytes)
        Interrupt: pin ? routed to IRQ 22

0000:00:0d.0 RAID bus controller: Promise Technology, Inc. 20267 (rev 02)
        Subsystem: Promise Technology, Inc.: Unknown device 4d32
        Control: I/O+ Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR- FastB2B-
        Status: Cap+ 66MHz- UDF- FastB2B- ParErr- DEVSEL=medium >TAbort- <TAbort- <MAbort- >SERR- <PERR-
        Latency: 32
        Interrupt: pin A routed to IRQ 23
        Region 0: I/O ports at 04f0 [size=80860000]
        Region 1: I/O ports at 04e0 [size=4]
        Region 2: I/O ports at 04d0 [size=8]
        Region 3: I/O ports at 04c0 [size=4]
        Region 4: I/O ports at 0480 [size=64]
        Region 5: Memory at 80840000 (32-bit, non-prefetchable) [size=128K]
        Expansion ROM at 00010000 [disabled]
        Capabilities: [58] Power Management version 1
                Flags: PMEClk- DSI- D1- D2- AuxCurrent=0mA PME(D0-,D1-,D2-,D3hot-,D3cold-)
                Status: D0 PME-Enable- DSel=0 DScale=0 PME-

.
.
0000:00:0f.0 Ethernet controller: Digital Equipment Corporation DECchip 21142/43 (rev 21)
        Control: I/O+ Mem+ BusMaster+ SpecCycle- MemWINV+ VGASnoop- ParErr- Stepping- SERR- FastB2B-
        Status: Cap- 66MHz- UDF- FastB2B+ ParErr- DEVSEL=medium >TAbort- <TAbort- <MAbort- >SERR- <PERR-
        Latency: 32 (5000ns min, 10000ns max), Cache Line Size: 0x08 (32 bytes)
        Interrupt: pin A routed to IRQ 25
        Region 0: I/O ports at 0400 [size=80800000]
        Region 1: Memory at 80871000 (32-bit, non-prefetchable) [size=128]
        Expansion ROM at 00040000 [disabled]

0000:00:10.0 Class ff00: Apple Computer Inc. Grand Central I/O (rev 02)
        Control: I/O- Mem+ BusMaster+ SpecCycle- MemWINV+ VGASnoop- ParErr- Stepping- SERR- FastB2B-
        Status: Cap- 66MHz- UDF- FastB2B- ParErr- DEVSEL=medium >TAbort- <TAbort- <MAbort- >SERR- <PERR+
        Latency: 32, Cache Line Size: 0x08 (32 bytes)
        Interrupt: pin ? routed to IRQ 22
        Region 0: Memory at f3000000 (32-bit, non-prefetchable)


Does this help at all?

Bob

Re: DMA appears broken in 2.6.11 for Mac 7200

[linuxppcdev]

Ben Said:
> On Mon, 2005-03-14 at 03:51 +0000, linuxppcdev@qbjnet.com wrote:
> > Just updated a couple of old Mac 7200 (601) boxen and ran into trouble with
> > the ide driver and a promise PDC20267 card. 
> > 
> > The very same drives and card work fine on a 9500 but hang on a 7200. If
> > I set the kernel arg ide=nodma I get in the dmesg:
> > ide_setup: ide=nodma : Prevented DMA
> > and the 7200 boots and runs fine.
> > 
> > Another suspect, the mace ethernet driver (which also appears to use DMA)
> > doesn't work on the 7200 with 2.6.11 but does with 2.4.28 and it works 
> > with 2.6.11 on the 9500.
> 
> Interesting. Not sure what's up, though. Those old machines were known
> to have bugs relative to cache coherency... also check wetehr we are
> setting the cache line size properly in PCI devices. You can also try to
> disable use of PCI memory write & invalidate command in all devices.
 
Did some more testing on this. Built a 2.4.30-pre1 kernel from bk source.

DMA works fine booting 2.4.30

I compared the dmesg and lspci and I see no differences in the cache line
size.

Is it possible that the ppc 601 kernel fixes aren't getting done properly
in 2.6??

Bob