riscv/vdso: make arch_setup_additional_pages wait for mmap_sem for write killable
authorTong Tiangen <tongtiangen@huawei.com>
Wed, 1 Sep 2021 02:46:21 +0000 (02:46 +0000)
committerPalmer Dabbelt <palmerdabbelt@google.com>
Sat, 2 Oct 2021 20:42:26 +0000 (13:42 -0700)
riscv architectures relying on mmap_sem for write in their
arch_setup_additional_pages. If the waiting task gets killed by the oom
killer it would block oom_reaper from asynchronous address space reclaim
and reduce the chances of timely OOM resolving.  Wait for the lock in
the killable mode and return with EINTR if the task got killed while
waiting.

Signed-off-by: Tong Tiangen <tongtiangen@huawei.com>
Reviewed-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Fixes: 76d2a0493a17 ("RISC-V: Init and Halt Code")
Signed-off-by: Palmer Dabbelt <palmerdabbelt@google.com>
arch/riscv/kernel/vdso.c

index e7bd92d..b70956d 100644 (file)
@@ -77,7 +77,9 @@ int arch_setup_additional_pages(struct linux_binprm *bprm,
 
        vdso_len = (vdso_pages + VVAR_NR_PAGES) << PAGE_SHIFT;
 
-       mmap_write_lock(mm);
+       if (mmap_write_lock_killable(mm))
+               return -EINTR;
+
        vdso_base = get_unmapped_area(NULL, 0, vdso_len, 0, 0);
        if (IS_ERR_VALUE(vdso_base)) {
                ret = vdso_base;