| Submitter | Myles Watson |
|---|---|
| Date | 2009-11-04 18:21:29 |
| Message ID | <2831fecf0911041021j1a32bdb1id4942a4254f647b@mail.gmail.com> |
| Download | mbox | patch |
| Permalink | /patch/534/ |
| State | Superseded |
| Headers | show |
Comments
Attached is the latest output. I don't see the errors anymore. I just tried booting Linux with the latest rom and it booted.
On Wed, Nov 4, 2009 at 2:48 PM, Hugh Greenberg <hng@lanl.gov> wrote: > Attached is the latest output. I don't see the errors anymore. I just > tried booting Linux with the latest rom and it booted. > Great! So is everything finally working, or are you still fighting sporadic hangs? We should commit the fixes so they don't get lost. Thanks, Myles
Myles, Everything seems to be working well. I had one sporadic hang while testing at what looked to be at the same place as the initial first hang. That was the only instance since I moved to gcc 3.4. We were testing so many things that I don't remember the exact scenario. I'm satisfied though. If I see it happening again, I'll try to capture it and send it to you. You deserve a medal for this. Thanks a lot.
On Wed, Nov 4, 2009 at 4:23 PM, Hugh Greenberg <hng@lanl.gov> wrote: > Myles, > > Everything seems to be working well. I had one sporadic hang while testing > at what looked to be at the same place as the initial first hang. That was > the only instance since I moved to gcc 3.4. We were testing so many things > that I don't remember the exact scenario. I'm satisfied though. If I see > it happening again, I'll try to capture it and send it to you. > Sounds good. I'd be interested in whether or not it hangs with the newer gcc still, too. Since everything else is working, it would be nice to know if that's really an issue. > You deserve a medal for this. Thanks a lot. No problem. Would you mind sending svn diff from your working tree so I can make a patch? I've lost track of which patches you're using. Thanks, Myles
Yes, it is still hanging occasionally with a newer gcc. I thought I put all the debug stuff back in, but I apparently didn't get it all. Below is the output. I will put the remaining debug statements back in for this hang again and send the output. coreboot-2.3 Wed Nov 4 16:40:23 MST 2009 starting... Enabling routing table for node 00 done. Enabling SMP settings (0,1) link=01 (1,0) link=01 setup_remote_node: done Renaming current temporary node to 01 done. Enabling routing table for node 01 done. 02 nodes initialized. coherent_ht_finalize done started ap apicid: SBLink=00 NC node|link=00 entering ht_optimize_link pos=0x8a, unfiltered freq_cap=0x8075 pos=0x8a, filtered freq_cap=0x35 pos=0xce, unfiltered freq_cap=0x35 freq_cap1=0x35, freq_cap2=0x15 dev1 old_freq=0x0, freq=0x4, needs_reset=0x1 dev2 old_freq=0x0, freq=0x4, needs_reset=0x1 width_cap1=0x11, width_cap2=0x11 dev1 input ln_width1=0x4, ln_width2=0x4 dev1 input width=0x1 dev1 output ln_width1=0x4, ln_width2=0x4 dev1 input|output width=0x11 old dev1 input|output width=0x11 dev2 input|output width=0x11 old dev2 input|output width=0x11 entering ht_optimize_link pos=0xd2, unfiltered freq_cap=0x35 pos=0xce, unfiltered freq_cap=0x1 pos=0xce, filtered freq_cap=0x1 freq_cap1=0x15, freq_cap2=0x1 dev1 old_freq=0x0, freq=0x0, needs_reset=0x0 dev2 old_freq=0x0, freq=0x0, needs_reset=0x0 width_cap1=0x0, width_cap2=0x0 dev1 input ln_width1=0x3, ln_width2=0x3 dev1 input width=0x0 dev1 output ln_width1=0x3, ln_width2=0x3 dev1 input|output width=0x0 old dev1 input|output width=0x0 dev2 input|output width=0x0 old dev2 input|output width=0x0 ht reset - coreboot-2.3 Wed Nov 4 16:40:23 MST 2009 starting... Enabling routing table for node 00 done. Enabling SMP settings (0,1) link=01 (1,0) link=01 setup_remote_node: done Renaming current temporary node to 01 done. Enabling routing table for node 01 done. 02 nodes initialized. coherent_ht_finalize done started ap apicid: SBLink=00 NC node|link=00 entering ht_optimize_link pos=0x8a, unfiltered freq_cap=0x8075 pos=0x8a, filtered freq_cap=0x35 pos=0xce, unfiltered freq_cap=0x35 freq_cap1=0x35, freq_cap2=0x15 dev1 old_freq=0x4, freq=0x4, needs_reset=0x0 dev2 old_freq=0x4, freq=0x4, needs_reset=0x0 width_cap1=0x11, width_cap2=0x11 dev1 input ln_width1=0x4, ln_width2=0x4 dev1 input width=0x1 dev1 output ln_width1=0x4, ln_width2=0x4 dev1 input|output width=0x11 old dev1 input|output width=0x11 dev2 input|output width=0x11 old dev2 input|output width=0x11 entering ht_optimize_link pos=0xd2, unfiltered freq_cap=0x35 pos=0xce, unfiltered freq_cap=0x1 pos=0xce, filtered freq_cap=0x1 freq_cap1=0x15, freq_cap2=0x1 dev1 old_freq=0x0, freq=0x0, needs_reset=0x0 dev2 old_freq=0x0, freq=0x0, needs_reset=0x0 width_cap1=0x0, width_cap2=0x0 dev1 input ln_width1=0x3, ln_width2=0x3 dev1 input width=0x0 dev1 output ln_width1=0x3, ln_width2=0x3 dev1 input|output width=0x0 old dev1 input|output width=0x0 dev2 input|output width=0x0 old dev2 input|output width=0x0 SMBus controller enabled Ram1.00 setting up CPU00 northbridge registers done. Ram1.01 setting up CPU01 northbridge registers done. Ram2.00 Enabling dual channel memory Registered 166Mhz RAM end at 0x00100000 kB Lower RAM end at 0x00100000 kB Ram2.01 Enabling dual channel memory Registered 166Mhz RAM end at 0x00200000 kB Lower RAM end at 0x00200000 kB Ram3
> Yes, it is still hanging occasionally with a newer gcc. Too bad. Occasionally like 1-in-3? I don't have any idea what the compiler would have to do with non-deterministic hangs. > I thought I put > all the debug stuff back in, but I apparently didn't get it all. Below > is the output. I will put the remaining debug statements back in for > this hang again and send the output. OK. The other thing to find out is if you still need the "Forcing Type 1" hack. Everything else is nearly trivial since it only affects your board, and you've tested it. Thanks, Myles
Sometimes it takes a very long time to reproduce. It is very random. I just tried about 30 times and I was not able to reproduce it. Ok, I will see if I still need that.
It doesn't seem like I need the forcing type 1 hack anymore. I reproduced the hang with some of the debugging statements back in. I hope that is all you need. Here is the output: coreboot-2.3 Wed Nov 4 17:28:13 MST 2009 starting... Enabling routing table for node 00 done. Enabling SMP settings (0,1) link=01 (1,0) link=01 setup_remote_node: done Renaming current temporary node to 01 done. Enabling routing table for node 01 done. 02 nodes initialized. coherent_ht_finalize done started ap apicid: SBLink=00 NC node|link=00 entering ht_optimize_link pos=0x8a, unfiltered freq_cap=0x8075 pos=0x8a, filtered freq_cap=0x35 pos=0xce, unfiltered freq_cap=0x35 freq_cap1=0x35, freq_cap2=0x15 dev1 old_freq=0x0, freq=0x4, needs_reset=0x1 dev2 old_freq=0x0, freq=0x4, needs_reset=0x1 width_cap1=0x11, width_cap2=0x11 dev1 input ln_width1=0x4, ln_width2=0x4 dev1 input width=0x1 dev1 output ln_width1=0x4, ln_width2=0x4 dev1 input|output width=0x11 old dev1 input|output width=0x11 dev2 input|output width=0x11 old dev2 input|output width=0x11 entering ht_optimize_link pos=0xd2, unfiltered freq_cap=0x35 pos=0xce, unfiltered freq_cap=0x1 pos=0xce, filtered freq_cap=0x1 freq_cap1=0x15, freq_cap2=0x1 dev1 old_freq=0x0, freq=0x0, needs_reset=0x0 dev2 old_freq=0x0, freq=0x0, needs_reset=0x0 width_cap1=0x0, width_cap2=0x0 dev1 input ln_width1=0x3, ln_width2=0x3 dev1 input width=0x0 dev1 output ln_width1=0x3, ln_width2=0x3 dev1 input|output width=0x0 old dev1 input|output width=0x0 dev2 input|output width=0x0 old dev2 input|output width=0x0 ht reset - coreboot-2.3 Wed Nov 4 17:28:13 MST 2009 starting... Enabling routing table for node 00 done. Enabling SMP settings (0,1) link=01 (1,0) link=01 setup_remote_node: done Renaming current temporary node to 01 done. Enabling routing table for node 01 done. 02 nodes initialized. coherent_ht_finalize done started ap apicid: SBLink=00 NC node|link=00 entering ht_optimize_link pos=0x8a, unfiltered freq_cap=0x8075 pos=0x8a, filtered freq_cap=0x35 pos=0xce, unfiltered freq_cap=0x35 freq_cap1=0x35, freq_cap2=0x15 dev1 old_freq=0x4, freq=0x4, needs_reset=0x0 dev2 old_freq=0x4, freq=0x4, needs_reset=0x0 width_cap1=0x11, width_cap2=0x11 dev1 input ln_width1=0x4, ln_width2=0x4 dev1 input width=0x1 dev1 output ln_width1=0x4, ln_width2=0x4 dev1 input|output width=0x11 old dev1 input|output width=0x11 dev2 input|output width=0x11 old dev2 input|output width=0x11 entering ht_optimize_link pos=0xd2, unfiltered freq_cap=0x35 pos=0xce, unfiltered freq_cap=0x1 pos=0xce, filtered freq_cap=0x1 freq_cap1=0x15, freq_cap2=0x1 dev1 old_freq=0x0, freq=0x0, needs_reset=0x0 dev2 old_freq=0x0, freq=0x0, needs_reset=0x0 width_cap1=0x0, width_cap2=0x0 dev1 input ln_width1=0x3, ln_width2=0x3 dev1 input width=0x0 dev1 output ln_width1=0x3, ln_width2=0x3 dev1 input|output width=0x0 old dev1 input|output width=0x0 dev2 input|output width=0x0 old dev2 input|output width=0x0 SMBus controller enabled Ram1.00 setting up CPU00 northbridge registers done. Ram1.01 setting up CPU01 northbridge registers done. Ram2.00 Enabling dual channel memory Registered 166Mhz RAM end at 0x00100000 kB Lower RAM end at 0x00100000 kB Ram2.01 Enabling dual channel memory Registered 166Mhz RAM end at 0x00200000 kB Lower RAM end at 0x00200000 kB Ram3 Before starting clocks: Before memreset:
> It doesn't seem like I need the forcing type 1 hack anymore. Good. > I reproduced the hang with some of the debugging statements back in. Thanks. > I hope that is all you need. Yeah. It took a little too long to figure that one out :) I think we'll call it good for now. > Here is the output: > Ram3 > Before starting clocks: Before memreset: I guess we should put the timeout back in so that it doesn't hang. I really don't understand why that would be compiler dependent. I'll send the patch to the list tomorrow. Thanks, Myles
Patch
Index: svn/src/mainboard/arima/hdama/mptable.c =================================================================== --- svn.orig/src/mainboard/arima/hdama/mptable.c +++ svn/src/mainboard/arima/hdama/mptable.c @@ -65,7 +65,7 @@ static unsigned node_link_to_bus(unsigne dev = dev_find_slot(0, PCI_DEVFN(0x18, 1)); if (!dev) { - return 0; + return 0xff; } for(reg = 0xE0; reg < 0xF0; reg += 0x04) { uint32_t config_map; @@ -89,7 +89,7 @@ static unsigned node_link_to_bus(unsigne return bus_base; } } - return 0; + return 0xff; } unsigned max_apicid(void) @@ -152,9 +152,9 @@ void *smp_write_config_table(void *v) /* HT chain 0 */ bus_chain_0 = node_link_to_bus(0, 0); - if (bus_chain_0 == 0) { + if (bus_chain_0 == 0xff) { printk_debug("ERROR - cound not find bus for node 0 chain 0, using defaults\n"); - bus_chain_0 = 1; + bus_chain_0 = 0; } /* 8111 */ @@ -165,7 +165,7 @@ void *smp_write_config_table(void *v) bus_isa++; } else { - printk_debug("ERROR - could not find PCI 1:03.0, using defaults\n"); + printk_debug("ERROR - could not find PCI %02x:03.0, using defaults\n", bus_chain_0); bus_8111_1 = 4; bus_isa = 5; @@ -177,7 +177,7 @@ void *smp_write_config_table(void *v) } else { - printk_debug("ERROR - could not find PCI 1:01.0, using defaults\n"); + printk_debug("ERROR - could not find PCI %02x:01.0, using defaults\n", bus_chain_0); bus_8131_1 = 2; } @@ -188,7 +188,7 @@ void *smp_write_config_table(void *v) } else { - printk_debug("ERROR - could not find PCI 1:02.0, using defaults\n"); + printk_debug("ERROR - could not find PCI %02x:02.0, using defaults\n", bus_chain_0); bus_8131_2 = 3; } @@ -327,16 +327,21 @@ void reboot_if_hotswap(void) int reset; int i; + unsigned char bus_chain_0 = node_link_to_bus(0, 0); reset = 0; printk_debug("Looking for bad PCIX MHz input\n"); - dev = dev_find_slot(1, PCI_DEVFN(0x02,0)); + dev = dev_find_slot(bus_chain_0, PCI_DEVFN(0x02,0)); + if (!dev) + printk_debug("Couldn't find %02x:02.0 \n", bus_chain_0); data = pci_read_config32(dev, 0xa0); if(!(((data>>16)&0x03)==0x03)) { reset=1; printk_debug("Bad PCIX MHz - Reset\n"); } printk_debug("Looking for bad Hot Swap Enable\n"); - dev = dev_find_slot(1, PCI_DEVFN(0x01,0)); + dev = dev_find_slot(bus_chain_0, PCI_DEVFN(0x01,0)); + if (!dev) + printk_debug("Couldn't find %02x:01.0 \n", bus_chain_0); data = pci_read_config32(dev, 0x48); if(data & 0x0c) { reset=1;