From 953f064aa6b29debcc211869b60bd59f26d19c34 Mon Sep 17 00:00:00 2001 From: Li Xinhai Date: Fri, 4 Sep 2020 16:36:10 -0700 Subject: mm/hugetlb: try preferred node first when alloc gigantic page from cma Since commit cf11e85fc08c ("mm: hugetlb: optionally allocate gigantic hugepages using cma"), the gigantic page would be allocated from node which is not the preferred node, although there are pages available from that node. The reason is that the nid parameter has been ignored in alloc_gigantic_page(). Besides, the __GFP_THISNODE also need be checked if user required to alloc only from the preferred node. After this patch, the preferred node is tried first before other allowed nodes, and don't try to allocate from other nodes if __GFP_THISNODE is specified. If user don't specify the preferred node, the current node will be used as preferred node, which makes sure consistent behavior of allocating gigantic and non-gigantic hugetlb page. Fixes: cf11e85fc08c ("mm: hugetlb: optionally allocate gigantic hugepages using cma") Signed-off-by: Li Xinhai Signed-off-by: Andrew Morton Reviewed-by: Mike Kravetz Acked-by: Michal Hocko Cc: Roman Gushchin Link: https://lkml.kernel.org/r/20200902025016.697260-1-lixinhai.lxh@gmail.com Signed-off-by: Linus Torvalds --- mm/hugetlb.c | 23 +++++++++++++++++------ 1 file changed, 17 insertions(+), 6 deletions(-) (limited to 'mm/hugetlb.c') diff --git a/mm/hugetlb.c b/mm/hugetlb.c index a301c2d672bf..5957dc80ebb1 100644 --- a/mm/hugetlb.c +++ b/mm/hugetlb.c @@ -1250,21 +1250,32 @@ static struct page *alloc_gigantic_page(struct hstate *h, gfp_t gfp_mask, int nid, nodemask_t *nodemask) { unsigned long nr_pages = 1UL << huge_page_order(h); + if (nid == NUMA_NO_NODE) + nid = numa_mem_id(); #ifdef CONFIG_CMA { struct page *page; int node; - for_each_node_mask(node, *nodemask) { - if (!hugetlb_cma[node]) - continue; - - page = cma_alloc(hugetlb_cma[node], nr_pages, - huge_page_order(h), true); + if (hugetlb_cma[nid]) { + page = cma_alloc(hugetlb_cma[nid], nr_pages, + huge_page_order(h), true); if (page) return page; } + + if (!(gfp_mask & __GFP_THISNODE)) { + for_each_node_mask(node, *nodemask) { + if (node == nid || !hugetlb_cma[node]) + continue; + + page = cma_alloc(hugetlb_cma[node], nr_pages, + huge_page_order(h), true); + if (page) + return page; + } + } } #endif -- cgit v1.2.3 From 17743798d81238ab13050e8e2833699b54e15467 Mon Sep 17 00:00:00 2001 From: Muchun Song Date: Fri, 4 Sep 2020 16:36:13 -0700 Subject: mm/hugetlb: fix a race between hugetlb sysctl handlers There is a race between the assignment of `table->data` and write value to the pointer of `table->data` in the __do_proc_doulongvec_minmax() on the other thread. CPU0: CPU1: proc_sys_write hugetlb_sysctl_handler proc_sys_call_handler hugetlb_sysctl_handler_common hugetlb_sysctl_handler table->data = &tmp; hugetlb_sysctl_handler_common table->data = &tmp; proc_doulongvec_minmax do_proc_doulongvec_minmax sysctl_head_finish __do_proc_doulongvec_minmax unuse_table i = table->data; *i = val; // corrupt CPU1's stack Fix this by duplicating the `table`, and only update the duplicate of it. And introduce a helper of proc_hugetlb_doulongvec_minmax() to simplify the code. The following oops was seen: BUG: kernel NULL pointer dereference, address: 0000000000000000 #PF: supervisor instruction fetch in kernel mode #PF: error_code(0x0010) - not-present page Code: Bad RIP value. ... Call Trace: ? set_max_huge_pages+0x3da/0x4f0 ? alloc_pool_huge_page+0x150/0x150 ? proc_doulongvec_minmax+0x46/0x60 ? hugetlb_sysctl_handler_common+0x1c7/0x200 ? nr_hugepages_store+0x20/0x20 ? copy_fd_bitmaps+0x170/0x170 ? hugetlb_sysctl_handler+0x1e/0x20 ? proc_sys_call_handler+0x2f1/0x300 ? unregister_sysctl_table+0xb0/0xb0 ? __fd_install+0x78/0x100 ? proc_sys_write+0x14/0x20 ? __vfs_write+0x4d/0x90 ? vfs_write+0xef/0x240 ? ksys_write+0xc0/0x160 ? __ia32_sys_read+0x50/0x50 ? __close_fd+0x129/0x150 ? __x64_sys_write+0x43/0x50 ? do_syscall_64+0x6c/0x200 ? entry_SYSCALL_64_after_hwframe+0x44/0xa9 Fixes: e5ff215941d5 ("hugetlb: multiple hstates for multiple page sizes") Signed-off-by: Muchun Song Signed-off-by: Andrew Morton Reviewed-by: Mike Kravetz Cc: Andi Kleen Link: http://lkml.kernel.org/r/20200828031146.43035-1-songmuchun@bytedance.com Signed-off-by: Linus Torvalds --- mm/hugetlb.c | 26 ++++++++++++++++++++------ 1 file changed, 20 insertions(+), 6 deletions(-) (limited to 'mm/hugetlb.c') diff --git a/mm/hugetlb.c b/mm/hugetlb.c index 5957dc80ebb1..67fc6383995b 100644 --- a/mm/hugetlb.c +++ b/mm/hugetlb.c @@ -3465,6 +3465,22 @@ static unsigned int allowed_mems_nr(struct hstate *h) } #ifdef CONFIG_SYSCTL +static int proc_hugetlb_doulongvec_minmax(struct ctl_table *table, int write, + void *buffer, size_t *length, + loff_t *ppos, unsigned long *out) +{ + struct ctl_table dup_table; + + /* + * In order to avoid races with __do_proc_doulongvec_minmax(), we + * can duplicate the @table and alter the duplicate of it. + */ + dup_table = *table; + dup_table.data = out; + + return proc_doulongvec_minmax(&dup_table, write, buffer, length, ppos); +} + static int hugetlb_sysctl_handler_common(bool obey_mempolicy, struct ctl_table *table, int write, void *buffer, size_t *length, loff_t *ppos) @@ -3476,9 +3492,8 @@ static int hugetlb_sysctl_handler_common(bool obey_mempolicy, if (!hugepages_supported()) return -EOPNOTSUPP; - table->data = &tmp; - table->maxlen = sizeof(unsigned long); - ret = proc_doulongvec_minmax(table, write, buffer, length, ppos); + ret = proc_hugetlb_doulongvec_minmax(table, write, buffer, length, ppos, + &tmp); if (ret) goto out; @@ -3521,9 +3536,8 @@ int hugetlb_overcommit_handler(struct ctl_table *table, int write, if (write && hstate_is_gigantic(h)) return -EINVAL; - table->data = &tmp; - table->maxlen = sizeof(unsigned long); - ret = proc_doulongvec_minmax(table, write, buffer, length, ppos); + ret = proc_hugetlb_doulongvec_minmax(table, write, buffer, length, ppos, + &tmp); if (ret) goto out; -- cgit v1.2.3