On Mon, Dec 06, 2004 at 09:43:16AM -0800, Dave Hansen wrote: > On Mon, 2004-12-06 at 09:28, Mark Wong wrote: > > I was trying out tests-041126 against the 2.6.10-rc1-mm5-mhp1 with > > page_section, local_memmap, local_memalloc, lru_drain_wq, nonlinear > > and nowriteback patches applied on top of that. I get the following > > oops when I run make test, let me know if I can provide more > > information: > ... > > EIP is at online_pages+0x7f/0xb0 > > eax: 2073096d ebx: 00000000 ecx: 2073096d edx: 2073096d > > esi: ffcae800 edi: 00008000 ebp: f747f3e0 esp: f750eea0 > > The odds are that is from this loop: > > for (i = 0; i < nr_pages; i++) > online_page(pfn_to_page(pfn + i)); > > So, either the pfn_to_page translation is bogus, or the pages aren't > mapped properly. I'd suspect the local_memmap patches. More complete > debugging info would include an 'addr2line -e c01669cf vmliunux' and > disassembly of online_page(). > I just tried 2.6.10-rc1-mm5-mhp1 with only the lru_drain_wq, nonlinear and nowriteback patches. When I trun make test I get a segfault: # make test ./run.sh 5 ./aioalloc online 8 (0x40000000) ./run.sh: line 21: 7069 Segmentation fault sh $r $i failed make: *** [test] Error 1 And the system gets a bit screwy after that, everything segfaults, connection drops... I don't see any errors on the console or in the logs. Is there another way I figure out what's going on? Mark
