From: "Zhangjian (Bamvor)" Subject: Re: [RFD] Efficient unit test and fuzz tools for kernel/libc porting Date: Thu, 21 Jul 2016 20:39:11 +0800 Message-ID: <5790C26F.4080408@huawei.com> References: <577CB5B7.7040204@huawei.com> Mime-Version: 1.0 Content-Type: text/plain; charset="windows-1252"; format=flowed Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: Sender: linux-kernel-owner@vger.kernel.org To: Joseph Myers Cc: LKML , Linux-Arch , libc-alpha@sourceware.org, trinity@vger.kernel.org, syzkaller , aponomarenko@rosalab.ru, Jess Hertz , Tim Newsham , Arnd Bergmann , Catalin Marinas , Mark Brown , Maxim Kuvyrkov , Yury Norov , Andrew Pinski , Andreas Schwab , Alexander Graf , marcus.shawcroft@arm.com, Ding Tianhong , Hanjun Guo , cuibixuan@huawei.com, lijinyue@huawei.com, Zefan Li List-Id: linux-arch.vger.kernel.org Hi, Joseph On 2016/7/20 23:47, Joseph Myers wrote: > On Wed, 6 Jul 2016, Zhangjian (Bamvor) wrote: > >> correct or not. After learn and compare some fuzz tools, I feel that there is >> no such fuzz tools could help me. So, I wrote a new fuzz tools base on the >> trinity and it found several wrapper issues in glibc. I will first explain the >> different with existing fuzz tools and paste my propsosal in the end. > > I'm not at all clear on whether any of the people working on AArch64 ILP32 > glibc have run the glibc testsuite and investigated the results in detail > (the patch submissions have failed to include glibc testsuite results and > have included bugs that would have been detected by the glibc testsuite). I run test glibc testsuite in previous glibc version with v6 kernel patch backport to kernel-4.1, without regression. I usually run glibc testsuite after ltp test result looks good. So, maybe it hard to find a issue by glibc testsuite in this case. > But, if you've found bugs in a new glibc port that were not detected by > the existing testsuite, then tests for those bugs should be contributed to > glibc (even if no existing port has those bugs, improving the test > coverage is still a good idea). It is good idea. I will review the fixed issues(such as wrong context in signal, wrong parameter in off_t/stat relative syscalls) and check if it is suitable to add it to glibc testsuite. (Actually, I do not know which test suite (ltp or glibc) I should improve for a specific issue). I hope our tools could help on improving the coverage of syscall relative code at least. Thanks. Bamvor