From f931ab479dd24cf7a2c6e2df19778406892591fb Mon Sep 17 00:00:00 2001 From: Dan Williams Date: Tue, 10 Jan 2017 16:57:36 -0800 Subject: [PATCH] mm: fix devm_memremap_pages crash, use mem_hotplug_{begin, done} Both arch_add_memory() and arch_remove_memory() expect a single threaded context. For example, arch/x86/mm/init_64.c::kernel_physical_mapping_init() does not hold any locks over this check and branch: if (pgd_val(*pgd)) { pud = (pud_t *)pgd_page_vaddr(*pgd); paddr_last = phys_pud_init(pud, __pa(vaddr), __pa(vaddr_end), page_size_mask); continue; } pud = alloc_low_page(); paddr_last = phys_pud_init(pud, __pa(vaddr), __pa(vaddr_end), page_size_mask); The result is that two threads calling devm_memremap_pages() simultaneously can end up colliding on pgd initialization. This leads to crash signatures like the following where the loser of the race initializes the wrong pgd entry: BUG: unable to handle kernel paging request at ffff888ebfff0000 IP: memcpy_erms+0x6/0x10 PGD 2f8e8fc067 PUD 0 /* <---- Invalid PUD */ Oops: 0000 [#1] SMP DEBUG_PAGEALLOC CPU: 54 PID: 3818 Comm: systemd-udevd Not tainted 4.6.7+ #13 task: ffff882fac290040 ti: ffff882f887a4000 task.ti: ffff882f887a4000 RIP: memcpy_erms+0x6/0x10 [..] Call Trace: ? pmem_do_bvec+0x205/0x370 [nd_pmem] ? blk_queue_enter+0x3a/0x280 pmem_rw_page+0x38/0x80 [nd_pmem] bdev_read_page+0x84/0xb0 Hold the standard memory hotplug mutex over calls to arch_{add,remove}_memory(). Fixes: 41e94a851304 ("add devm_memremap_pages") Link: http://lkml.kernel.org/r/148357647831.9498.12606007370121652979.stgit@dwillia2-desk3.amr.corp.intel.com Signed-off-by: Dan Williams Cc: Christoph Hellwig Cc: Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- kernel/memremap.c | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/kernel/memremap.c b/kernel/memremap.c index b501e390bb3440..9ecedc28b928de 100644 --- a/kernel/memremap.c +++ b/kernel/memremap.c @@ -246,7 +246,9 @@ static void devm_memremap_pages_release(struct device *dev, void *data) /* pages are dead and unused, undo the arch mapping */ align_start = res->start & ~(SECTION_SIZE - 1); align_size = ALIGN(resource_size(res), SECTION_SIZE); + mem_hotplug_begin(); arch_remove_memory(align_start, align_size); + mem_hotplug_done(); untrack_pfn(NULL, PHYS_PFN(align_start), align_size); pgmap_radix_release(res); dev_WARN_ONCE(dev, pgmap->altmap && pgmap->altmap->alloc, @@ -358,7 +360,9 @@ void *devm_memremap_pages(struct device *dev, struct resource *res, if (error) goto err_pfn_remap; + mem_hotplug_begin(); error = arch_add_memory(nid, align_start, align_size, true); + mem_hotplug_done(); if (error) goto err_add_memory;