Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Sat, 7 Aug 2021 11:45:45 -0700
From:   Andrew Morton <akpm@linux-foundation.org>
To:     Miaohe Lin <linmiaohe@huawei.com>, <imbrenda@linux.ibm.com>,
        <kirill.shutemov@linux.intel.com>, <jack@suse.cz>,
        <jhubbard@nvidia.com>, <linux-mm@kvack.org>,
        <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH 4/5] mm: gup: fix potential pgmap refcnt leak in
 __gup_device_huge()
Message-Id: <20210807114545.eaee318909a9215c4e056aa3@linux-foundation.org>
In-Reply-To: <20210807114112.6e45b31c65dd62169fee8718@linux-foundation.org>
References: <20210807093620.21347-1-linmiaohe@huawei.com>
        <20210807093620.21347-5-linmiaohe@huawei.com>
        <20210807114112.6e45b31c65dd62169fee8718@linux-foundation.org>
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Precedence: bulk

On Sat, 7 Aug 2021 11:41:12 -0700 Andrew Morton <akpm@linux-foundation.org> wrote:

> We can simplify further, and remove the troublesome multiple return points?
> 

oops.

--- a/mm/gup.c~mm-gup-fix-potential-pgmap-refcnt-leak-in-__gup_device_huge-fix-fix
+++ a/mm/gup.c
@@ -2240,6 +2240,7 @@ static int __gup_device_huge(unsigned lo
 {
 	int nr_start = *nr;
 	struct dev_pagemap *pgmap = NULL;
+	int ret = 1;
 
 	do {
 		struct page *page = pfn_to_page(pfn);
@@ -2247,12 +2248,14 @@ static int __gup_device_huge(unsigned lo
 		pgmap = get_dev_pagemap(pfn, pgmap);
 		if (unlikely(!pgmap)) {
 			undo_dev_pagemap(nr, nr_start, flags, pages);
+			ret = 0;
 			break;
 		}
 		SetPageReferenced(page);
 		pages[*nr] = page;
 		if (unlikely(!try_grab_page(page, flags))) {
 			undo_dev_pagemap(nr, nr_start, flags, pages);
+			ret = 0;
 			break;
 		}
 		(*nr)++;
@@ -2260,7 +2263,7 @@ static int __gup_device_huge(unsigned lo
 	} while (addr += PAGE_SIZE, addr != end);
 
 	put_dev_pagemap(pgmap);
-	return 1;
+	return ret;
 }
 
 static int __gup_device_huge_pmd(pmd_t orig, pmd_t *pmdp, unsigned long addr,

Not sure if it's worth bothering, really...