diff options
Diffstat (limited to 'results/classifier/118/all/1716292')
| -rw-r--r-- | results/classifier/118/all/1716292 | 92 |
1 files changed, 92 insertions, 0 deletions
diff --git a/results/classifier/118/all/1716292 b/results/classifier/118/all/1716292 new file mode 100644 index 000000000..f6f4a78d3 --- /dev/null +++ b/results/classifier/118/all/1716292 @@ -0,0 +1,92 @@ +network: 0.952 +virtual: 0.951 +semantic: 0.951 +debug: 0.949 +PID: 0.945 +device: 0.943 +register: 0.942 +permissions: 0.940 +ppc: 0.940 +graphic: 0.937 +arm: 0.936 +performance: 0.936 +assembly: 0.935 +user-level: 0.934 +files: 0.933 +vnc: 0.933 +architecture: 0.932 +TCG: 0.923 +socket: 0.922 +risc-v: 0.917 +x86: 0.897 +hypervisor: 0.896 +mistranslation: 0.884 +VMM: 0.881 +peripherals: 0.877 +boot: 0.869 +KVM: 0.841 +kernel: 0.833 +i386: 0.707 + +User mode emulation returns wrong value for write(fd, NULL, 0) + +QEMU version: latest master (fcea73709b966a7ded9efa7b106ea50c7fe9025c) +OS version: Ubuntu 14.04.3 +Configured with: ../configure --target-list=x86_64-linux-user + +QEMU Linux usermode emulation does not handle write() syscalls with zero length and a null pointer correctly: on Linux this returns 0, but in emulation this returns -1. + +I ran into this while using an aarch64 abuild-tar from Alpine Linux in user-mode emulation; here's the minimized reproduction test case: + +zhuowei@zhuowei-tablet:/tmp$ cat writezerobytes.c +#include <stdio.h> +#include <unistd.h> +#include <fcntl.h> + +int main() { + ssize_t ret = write(STDOUT_FILENO, NULL, 0); + fprintf(stderr, "write returned %ld\n", ret); + return 0; +} +zhuowei@zhuowei-tablet:/tmp$ gcc -o writezerobytes writezerobytes.c +zhuowei@zhuowei-tablet:/tmp$ uname -a +Linux zhuowei-tablet 3.13.0-129-generic #178-Ubuntu SMP Fri Aug 11 12:48:20 UTC 2017 x86_64 x86_64 x86_64 GNU/Linux +zhuowei@zhuowei-tablet:/tmp$ ./writezerobytes +write returned 0 +zhuowei@zhuowei-tablet:/tmp$ /media/zhuowei/redhd/docs/repos/qemu/build4/x86_64-linux-user/qemu-x86_64 ./writezerobytes +write returned -1 +zhuowei@zhuowei-tablet:/tmp$ /media/zhuowei/redhd/docs/repos/qemu/build4/x86_64-linux-user/qemu-x86_64 --version +qemu-x86_64 version 2.10.50 (v2.10.0-471-gfcea737-dirty) +Copyright (c) 2003-2017 Fabrice Bellard and the QEMU Project developers + +This happens for me also, with qemu version 2.12.0 (Debian 1:2.12+dfsg-3). + +An initial patch was proposed here: https://lists.gnu.org/archive/html/qemu-devel/2017-09/msg08073.html + +Discussion pointed out some problems, and the patch languished and was not accepted. + +Here is a summary of the changes needed for it to be more likely for the patch to be accepted: https://lists.gnu.org/archive/html/qemu-devel/2018-02/msg03964.html + + - change from "ret = 0" to something like "ret = get_errno(safe_write(arg1, NULL, 0))" + - change TARGET_NR_read to do the same, instead of its current short-circuit behaviour for count==0 + - check pread64/pwrite64 to see if they need a similar change as well + + + +On 09/07/2018 06:51 AM, Tony Garnock-Jones wrote: +> ** Patch added: "0001-Bring-linux-user-write-2-handling-into-line-with-lin.patch" +> https://bugs.launchpad.net/qemu/+bug/1716292/+attachment/5186008/+files/0001-Bring-linux-user-write-2-handling-into-line-with-lin.patch + +While a developer can chase a URL, our CI tools can't. Can you please +also send that patch directly to <email address hidden>, so that it gets +the same level of review as other patches? + +-- +Eric Blake, Principal Software Engineer +Red Hat, Inc. +1-919-301-3266 +Virtualization: qemu.org | libvirt.org + + +Fix has been committed here: +https://git.qemu.org/?p=qemu.git;a=commitdiff;h=58cfa6c2e6eb51b23cc98 + |