lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
Open Source and information security mailing list archives
| ||
|
Date: Tue, 28 Apr 2020 22:52:57 +0800 From: Dejin Zheng <zhengdejin5@...il.com> To: Andy Shevchenko <andy.shevchenko@...il.com> Cc: Nicolas Ferre <nicolas.ferre@...rochip.com>, "David S. Miller" <davem@...emloft.net>, Paul Walmsley <paul.walmsley@...ive.com>, Palmer Dabbelt <palmer@...belt.com>, yash.shah@...ive.com, netdev <netdev@...r.kernel.org>, Linux Kernel Mailing List <linux-kernel@...r.kernel.org>, Claudiu Beznea <Claudiu.Beznea@...rochip.com> Subject: Re: [PATCH net v1] net: macb: fix an issue about leak related system resources On Tue, Apr 28, 2020 at 05:04:32PM +0300, Andy Shevchenko wrote: > On Tue, Apr 28, 2020 at 4:12 PM Dejin Zheng <zhengdejin5@...il.com> wrote: > > On Tue, Apr 28, 2020 at 10:42:56AM +0200, Nicolas Ferre wrote: > > > On 28/04/2020 at 05:24, Dejin Zheng wrote: > > > > On Mon, Apr 27, 2020 at 01:33:41PM +0300, Andy Shevchenko wrote: > > > > > On Sat, Apr 25, 2020 at 3:57 PM Dejin Zheng <zhengdejin5@...il.com> wrote: > > > > > > > > > > > > A call of the function macb_init() can fail in the function > > > > > > fu540_c000_init. The related system resources were not released > > > > > > then. use devm_ioremap() to replace ioremap() for fix it. > > > > > > > > > > > > > > > > Why not to go further and convert to use devm_platform_ioremap_resource()? > > > > > > > > > devm_platform_ioremap_resource() will call devm_request_mem_region(), > > > > and here did not do it. > > > > > > And what about devm_platform_get_and_ioremap_resource()? This would > > > streamline this whole fu540_c000_init() function. > > > > > Nicolas, the function devm_platform_get_and_ioremap_resource() will also > > call devm_request_mem_region(), after call it, These IO addresses will > > be monopolized by this driver. the devm_ioremap() and ioremap() are not > > do this. if this IO addresses will be shared with the other driver, call > > devm_platform_get_and_ioremap_resource() may be fail. > > I guess request region is a right thing to do. If driver is sharing > this IO region with something else, it is a delayed bomb attack and > has to be fixed. > > I did encounter IO address sharing, for example, some registers are shared by both PHY and USB controllers on Tegra SoCs. See link[1] for details. If many people think that devm_request_mem_region() should be added here, I will send patch v2 and convert to use devm_platform_ioremap_resource(). Thanks very much! link[1]: https://patchwork.ozlabs.org/project/linux-tegra/patch/20200127135841.17935-1-zhengdejin5@gmail.com/ BR, Dejin > > > -- > With Best Regards, > Andy Shevchenko
Powered by blists - more mailing lists