From: Zhang Jian zhangjian210@huawei.com
hulk inclusion category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/I54IL8 CVE: NA backport: openEuler-22.03-LTS
-----------------------------
When passing numa id to sp_alloc, sometimes numa id does not work. This is because memory policy will change numa id to a preferred one if memory policy is set. Fix the error by mbind virtual address to desired numa id.
Signed-off-by: Zhang Jian zhangjian210@huawei.com Signed-off-by: Zhou Guanghui zhouguanghui1@huawei.com Signed-off-by: Wang Wensheng wangwensheng4@huawei.com Reviewed-by: Weilong Chen chenweilong@huawei.com Reviewed-by: Kefeng Wang wangkefeng.wang@huawei.com Signed-off-by: Zheng Zengkai zhengzengkai@huawei.com --- include/linux/mempolicy.h | 10 ++++++ mm/mempolicy.c | 14 ++++++--- mm/share_pool.c | 66 +++++++++++++++++++++++++-------------- 3 files changed, 62 insertions(+), 28 deletions(-)
diff --git a/include/linux/mempolicy.h b/include/linux/mempolicy.h index 64ab4398ba90..133a8b5bdf9e 100644 --- a/include/linux/mempolicy.h +++ b/include/linux/mempolicy.h @@ -201,6 +201,9 @@ extern bool vma_migratable(struct vm_area_struct *vma); extern int mpol_misplaced(struct page *, struct vm_area_struct *, unsigned long); extern void mpol_put_task_policy(struct task_struct *);
+extern long __do_mbind(unsigned long start, unsigned long len, + unsigned short mode, unsigned short mode_flags, + nodemask_t *nmask, unsigned long flags, struct mm_struct *mm); #else
struct mempolicy {}; @@ -301,6 +304,13 @@ static inline int mpol_misplaced(struct page *page, struct vm_area_struct *vma, return -1; /* no node preference */ }
+static long __do_mbind(unsigned long start, unsigned long len, + unsigned short mode, unsigned short mode_flags, + nodemask_t *nmask, unsigned long flags, struct mm_struct *mm) +{ + return 0; +} + static inline void mpol_put_task_policy(struct task_struct *task) { } diff --git a/mm/mempolicy.c b/mm/mempolicy.c index 5ce39dbc84e1..d2326f9a38a8 100644 --- a/mm/mempolicy.c +++ b/mm/mempolicy.c @@ -1308,11 +1308,10 @@ static struct page *new_page(struct page *page, unsigned long start) } #endif
-static long do_mbind(unsigned long start, unsigned long len, - unsigned short mode, unsigned short mode_flags, - nodemask_t *nmask, unsigned long flags) +long __do_mbind(unsigned long start, unsigned long len, + unsigned short mode, unsigned short mode_flags, + nodemask_t *nmask, unsigned long flags, struct mm_struct *mm) { - struct mm_struct *mm = current->mm; struct mempolicy *new; unsigned long end; int err; @@ -1411,6 +1410,13 @@ static long do_mbind(unsigned long start, unsigned long len, return err; }
+static long do_mbind(unsigned long start, unsigned long len, + unsigned short mode, unsigned short mode_flags, + nodemask_t *nmask, unsigned long flags) +{ + return __do_mbind(start, len, mode, mode_flags, nmask, flags, current->mm); +} + /* * User space interface with variable sized bitmaps for nodelists. */ diff --git a/mm/share_pool.c b/mm/share_pool.c index 85d175def6ae..76088952d0a5 100644 --- a/mm/share_pool.c +++ b/mm/share_pool.c @@ -16,7 +16,6 @@ * it under the terms of the GNU General Public License version 2 as * published by the Free Software Foundation. */ - #define pr_fmt(fmt) "share pool: " fmt
#include <linux/share_pool.h> @@ -2157,6 +2156,7 @@ struct sp_alloc_context { bool need_fallocate; struct timespec64 start; struct timespec64 end; + bool have_mbind; };
static void trace_sp_alloc_begin(struct sp_alloc_context *ac) @@ -2298,6 +2298,7 @@ static int sp_alloc_prepare(unsigned long size, unsigned long sp_flags, ac->sp_flags = sp_flags; ac->state = ALLOC_NORMAL; ac->need_fallocate = false; + ac->have_mbind = false; return 0; }
@@ -2391,7 +2392,7 @@ static void sp_alloc_fallback(struct sp_area *spa, struct sp_alloc_context *ac) }
static int sp_alloc_populate(struct mm_struct *mm, struct sp_area *spa, - struct sp_group_node *spg_node, struct sp_alloc_context *ac) + struct sp_alloc_context *ac) { int ret = 0; unsigned long sp_addr = spa->va_start; @@ -2423,25 +2424,20 @@ static int sp_alloc_populate(struct mm_struct *mm, struct sp_area *spa, if (ret) sp_add_work_compact(); } - if (ret) { - if (spa->spg != spg_none) - sp_alloc_unmap(list_next_entry(spg_node, proc_node)->master->mm, spa, spg_node); - else - sp_munmap(mm, spa->va_start, spa->real_size); - - if (unlikely(fatal_signal_pending(current))) - pr_warn_ratelimited("allocation failed, current thread is killed\n"); - else - pr_warn_ratelimited("allocation failed due to mm populate failed(potential no enough memory when -12): %d\n", - ret); - sp_fallocate(spa); /* need this, otherwise memleak */ - sp_alloc_fallback(spa, ac); - } else { - ac->need_fallocate = true; - } return ret; }
+static long sp_mbind(struct mm_struct *mm, unsigned long start, unsigned long len, + unsigned long node) +{ + nodemask_t nmask; + + nodes_clear(nmask); + node_set(node, nmask); + return __do_mbind(start, len, MPOL_BIND, MPOL_F_STATIC_NODES, + &nmask, MPOL_MF_STRICT, mm); +} + static int __sp_alloc_mmap_populate(struct mm_struct *mm, struct sp_area *spa, struct sp_group_node *spg_node, struct sp_alloc_context *ac) { @@ -2457,7 +2453,34 @@ static int __sp_alloc_mmap_populate(struct mm_struct *mm, struct sp_area *spa, return ret; }
- ret = sp_alloc_populate(mm, spa, spg_node, ac); + if (!ac->have_mbind) { + ret = sp_mbind(mm, spa->va_start, spa->real_size, spa->node_id); + if (ret < 0) { + pr_err("cannot bind the memory range to specified node:%d, err:%d\n", + spa->node_id, ret); + goto err; + } + ac->have_mbind = true; + } + + ret = sp_alloc_populate(mm, spa, ac); + if (ret) { +err: + if (spa->spg != spg_none) + sp_alloc_unmap(list_next_entry(spg_node, proc_node)->master->mm, spa, spg_node); + else + sp_munmap(mm, spa->va_start, spa->real_size); + + if (unlikely(fatal_signal_pending(current))) + pr_warn_ratelimited("allocation failed, current thread is killed\n"); + else + pr_warn_ratelimited("allocation failed due to mm populate failed(potential no enough memory when -12): %d\n", + ret); + sp_fallocate(spa); /* need this, otherwise memleak */ + sp_alloc_fallback(spa, ac); + } else + ac->need_fallocate = true; + return ret; }
@@ -2479,11 +2502,6 @@ static int sp_alloc_mmap_populate(struct sp_area *spa, if (mmap_ret) { if (ac->state != ALLOC_COREDUMP) return mmap_ret; - if (ac->spg == spg_none) { - sp_alloc_unmap(mm, spa, spg_node); - pr_err("dvpp allocation failed due to coredump"); - return mmap_ret; - } ac->state = ALLOC_NORMAL; continue; }