From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jann Horn Date: Tue, 28 Apr 2020 03:27:45 +0000 Subject: [PATCH 5/5] mm/gup: Take mmap_sem in get_dump_page() Message-Id: <20200428032745.133556-6-jannh@google.com> List-Id: References: <20200428032745.133556-1-jannh@google.com> In-Reply-To: <20200428032745.133556-1-jannh@google.com> MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: Andrew Morton Cc: Rich Felker , linux-c6x-dev@linux-c6x.org, Yoshinori Sato , linux-sh@vger.kernel.org, linux-kernel@vger.kernel.org, Oleg Nesterov , linux-mm@kvack.org, Alexander Viro , Mark Salter , linux-fsdevel@vger.kernel.org, Russell King , Aurelien Jacquiot , Linus Torvalds , Christoph Hellwig , linux-arm-kernel@lists.infradead.org, "Eric W . Biederman" Properly take the mmap_sem before calling into the GUP code from get_dump_page(); and play nice, allowing __get_user_pages_locked() to drop the mmap_sem if it has to sleep. This requires adjusting the check in __get_user_pages_locked() to be slightly less strict: While `vmas != NULL` is normally incompatible with the lock-dropping retry logic, it's fine if we only want a single page, because then retries can only happen when we haven't grabbed any pages yet. Signed-off-by: Jann Horn --- mm/gup.c | 21 ++++++++++++++++----- 1 file changed, 16 insertions(+), 5 deletions(-) diff --git a/mm/gup.c b/mm/gup.c index 9a7e83772f1fe..4bb4149c0e259 100644 --- a/mm/gup.c +++ b/mm/gup.c @@ -1261,7 +1261,8 @@ static __always_inline long __get_user_pages_locked(struct task_struct *tsk, if (locked) { /* if VM_FAULT_RETRY can be returned, vmas become invalid */ - BUG_ON(vmas); + if (WARN_ON(vmas && nr_pages != 1)) + return -EFAULT; /* check caller initialized locked */ BUG_ON(*locked != 1); } @@ -1548,18 +1549,28 @@ static long __get_user_pages_locked(struct task_struct *tsk, * NULL wherever the ZERO_PAGE, or an anonymous pte_none, has been found - * allowing a hole to be left in the corefile to save diskspace. * - * Called without mmap_sem, but after all other threads have been killed. + * Called without mmap_sem (takes and releases the mmap_sem by itself). */ struct page *get_dump_page(unsigned long addr) { + struct mm_struct *mm = current->mm; struct vm_area_struct *vma; struct page *page; + int locked = 1; + int ret; - if (__get_user_pages(current, current->mm, addr, 1, - FOLL_FORCE | FOLL_DUMP | FOLL_GET, &page, &vma, - NULL) < 1) + if (down_read_killable(&mm->mmap_sem)) + return NULL; + ret = __get_user_pages_locked(current, mm, addr, 1, &page, &vma, + &locked, + FOLL_FORCE | FOLL_DUMP | FOLL_GET); + if (ret != 1) { + if (locked) + up_read(&mm->mmap_sem); return NULL; + } flush_cache_page(vma, addr, page_to_pfn(page)); + up_read(&mm->mmap_sem); return page; } -- 2.26.2.303.gf8c07b1a785-goog