Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Content-Type: text/plain; charset="us-ascii"
Content-Disposition: inline
Date:   Wed, 10 Apr 2019 09:18:50 +0800
From:   Huang Shijie <sjhuang@iluvatar.ai>
To:     Ira Weiny <ira.weiny@intel.com>
CC:     Matthew Wilcox <willy@infradead.org>, <akpm@linux-foundation.org>,
        <william.kucharski@oracle.com>, <palmer@sifive.com>,
        <axboe@kernel.dk>, <keescook@chromium.org>, <linux-mm@kvack.org>,
        <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH 1/2] mm/gup.c: fix the wrong comments
Message-ID: <20190410011850.GA3640@hsj-Precision-5520>
References: <20190408023746.16916-1-sjhuang@iluvatar.ai>
 <20190408141313.GU22763@bombadil.infradead.org>
 <20190409010832.GA28081@hsj-Precision-5520>
 <20190409202316.GA22989@iweiny-DESK2.sc.intel.com>
MIME-Version: 1.0
In-Reply-To: <20190409202316.GA22989@iweiny-DESK2.sc.intel.com>
User-Agent: Mutt/1.9.4 (2018-02-28)
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Tue, Apr 09, 2019 at 01:23:16PM -0700, Ira Weiny wrote:
> On Tue, Apr 09, 2019 at 09:08:33AM +0800, Huang Shijie wrote:
> > On Mon, Apr 08, 2019 at 07:13:13AM -0700, Matthew Wilcox wrote:
> > > On Mon, Apr 08, 2019 at 10:37:45AM +0800, Huang Shijie wrote:
> > > > When CONFIG_HAVE_GENERIC_GUP is defined, the kernel will use its own
> > > > get_user_pages_fast().
> > > > 
> > > > In the following scenario, we will may meet the bug in the DMA case:
> > > > 	    .....................
> > > > 	    get_user_pages_fast(start,,, pages);
> > > > 	        ......
> > > > 	    sg_alloc_table_from_pages(, pages, ...);
> > > > 	    .....................
> > > > 
> > > > The root cause is that sg_alloc_table_from_pages() requires the
> > > > page order to keep the same as it used in the user space, but
> > > > get_user_pages_fast() will mess it up.
> > > 
> > > I don't understand how get_user_pages_fast() can return the pages in a
> > > different order in the array from the order they appear in userspace.
> > > Can you explain?
> > Please see the code in gup.c:
> > 
> > 	int get_user_pages_fast(unsigned long start, int nr_pages,
> > 				unsigned int gup_flags, struct page **pages)
> > 	{
> > 		.......
> > 		if (gup_fast_permitted(start, nr_pages)) {
> > 			local_irq_disable();
> > 			gup_pgd_range(addr, end, gup_flags, pages, &nr);               // The @pages array maybe filled at the first time.
> > 			local_irq_enable();
> > 			ret = nr;
> > 		}
> > 		.......
> > 		if (nr < nr_pages) {
> > 			/* Try to get the remaining pages with get_user_pages */
> > 			start += nr << PAGE_SHIFT;
> > 			pages += nr;                                                  // The @pages is moved forward.
> > 
> > 			if (gup_flags & FOLL_LONGTERM) {
> > 				down_read(&current->mm->mmap_sem);
> > 				ret = __gup_longterm_locked(current, current->mm,      // The @pages maybe filled at the second time
> >
> 
> Neither this nor the get_user_pages_unlocked is filling the pages a second
The get_user_pages_unlocked() will call the handle_mm_fault which will allocate a
new page for the empty PTE, and save the new page into the @pages array.


> time.  It is adding to the page array having moved start and the page array
> forward.

Yes. This will mess up the page order.

I will read the code again to check if I am wrong :)

> 
> Are you doing a FOLL_LONGTERM GUP?  Or are you in the else clause below when
> you get this bug?
I do not use FOLL_LONGTERM, I just use the FOLL_WRITE.

So it seems it runs into the else clause below.

Thanks
Huang Shijie

> 
> Ira
> 
> > 							    start, nr_pages - nr,
> > 							    pages, NULL, gup_flags);
> > 				up_read(&current->mm->mmap_sem);
> > 			} else {
> > 				/*
> > 				 * retain FAULT_FOLL_ALLOW_RETRY optimization if
> > 				 * possible
> > 				 */
> > 				ret = get_user_pages_unlocked(start, nr_pages - nr,    // The @pages maybe filled at the second time.
> > 							      pages, gup_flags);
> > 			}
> > 		}
> > 
> >