On Friday 24 January 2014 01:38 AM, Santosh Shilimkar wrote: > Yinghai, > > On Friday 24 January 2014 12:55 AM, Yinghai Lu wrote: >> On Thu, Jan 23, 2014 at 2:49 PM, Dave Hansen <dave.hansen@xxxxxxxxx> wrote: >>>> Linus's current tree doesn't boot on an 8-node/1TB NUMA system that I >>>> have. Its reboots are *LONG*, so I haven't fully bisected it, but it's >>>> down to a just a few commits, most of which are changes to the memblock >>>> code. Since the panic is in the memblock code, it looks like a >>>> no-brainer. It's almost certainly the code from Santosh or Grygorii >>>> that's triggering this. >>>> >>>> Config and good/bad dmesg with memblock=debug are here: >>>> >>>> http://sr71.net/~dave/intel/3.13/ >>>> >>>> Please let me know if you need it bisected further than this. >> Please check attached patch, and it should fix the problem. >> > > [...] > >> >> Subject: [PATCH] x86: Fix numa with reverting wrong memblock setting. >> >> Dave reported Numa on x86 is broken on system with 1T memory. >> >> It turns out >> | commit 5b6e529521d35e1bcaa0fe43456d1bbb335cae5d >> | Author: Santosh Shilimkar <santosh.shilimkar@xxxxxx> >> | Date: Tue Jan 21 15:50:03 2014 -0800 >> | >> | x86: memblock: set current limit to max low memory address >> >> set limit to low wrongly. >> >> max_low_pfn_mapped is different from max_pfn_mapped. >> max_low_pfn_mapped is always under 4G. >> >> That will memblock_alloc_nid all go under 4G. >> >> Revert that offending patch. >> >> Reported-by: Dave Hansen <dave.hansen@xxxxxxxxx> >> Signed-off-by: Yinghai Lu <yinghai@xxxxxxxxxx> >> >> > This mostly will fix the $subject issue but the regression > reported by Andrew [1] will surface with the revert. Its clear > now that even though commit fixed the issue, it wasn't the fix. > > Would be great if you can have a look at the thread. > The patch which is now commit 457ff1d {lib/swiotlb.c: use memblock apis for early memory allocations} was the breaking the boot on Andrew's machine. Now if I look back the patch, based on your above description, I believe below hunk waS/is the culprit. @@ -172,8 +172,9 @@ int __init swiotlb_init_with_tbl(char *tlb, unsigned long nslabs, int verbose) /* * Get the overflow emergency buffer */ - v_overflow_buffer = alloc_bootmem_low_pages_nopanic( - PAGE_ALIGN(io_tlb_overflow)); + v_overflow_buffer = memblock_virt_alloc_nopanic( + PAGE_ALIGN(io_tlb_overflow), + PAGE_SIZE); if (!v_overflow_buffer) return -ENOMEM; Looks like 'v_overflow_buffer' must be allocated from low memory in this case. Is that correct ? Regards, Santosh -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>