From: Zhangjin Wu <falcon@tinylab.org> To: thomas@t-8ch.de, w@1wt.eu Cc: falcon@tinylab.org, linux-kernel@vger.kernel.org, linux-kselftest@vger.kernel.org, linux-riscv@lists.infradead.org, palmer@dabbelt.com, paul.walmsley@sifive.com Subject: Re: [PATCH 13/13] tools/nolibc: sys_gettimeofday: riscv: use __NR_clock_gettime64 for rv32 Date: Sat, 27 May 2023 09:26:35 +0800 [thread overview] Message-ID: <20230527012635.19595-1-falcon@tinylab.org> (raw) In-Reply-To: <29e4b23d-27cb-4101-bfe9-c6b412222578@t-8ch.de> Hi, Thomas, Willy > On 2023-05-25 02:03:32+0800, Zhangjin Wu wrote: > > rv32 uses the generic include/uapi/asm-generic/unistd.h and it has no > > __NR_gettimeofday and __NR_clock_gettime after kernel commit d4c08b9776b3 > > ("riscv: Use latest system call ABI"), use __NR_clock_gettime64 instead. > > > > This code is based on src/time/gettimeofday.c of musl and > > sysdeps/unix/sysv/linux/clock_gettime.c of glibc. > > > > Both __NR_clock_gettime and __NR_clock_gettime64 are added for > > sys_gettimeofday() for they share most of the code. > > > > Notes: > > > > * Both tv and tz are not directly passed to kernel clock_gettime* > > syscalls, so, it isn't able to check the pointer automatically with the > > get_user/put_user helpers just like kernel gettimeofday syscall does. > > instead, we emulate (but not completely) such checks in our new > > __NR_clock_gettime* branch of nolibc. > > > > * kernel clock_gettime* syscalls can not get tz info, just like musl and > > glibc do, we set tz to zero to avoid a random number. > > > > Signed-off-by: Zhangjin Wu <falcon@tinylab.org> > > --- > > tools/include/nolibc/sys.h | 46 ++++++++++++++++++++++++++++++++++++++ > > 1 file changed, 46 insertions(+) > > > > diff --git a/tools/include/nolibc/sys.h b/tools/include/nolibc/sys.h > > index 2642b380c6aa..ad38cc3856be 100644 > > --- a/tools/include/nolibc/sys.h > > +++ b/tools/include/nolibc/sys.h > > @@ -26,6 +26,7 @@ > > > > #include "arch.h" > > #include "errno.h" > > +#include "string.h" > > #include "types.h" > > > > > > @@ -51,6 +52,11 @@ > > * should not be placed here. > > */ > > > > +/* > > + * This is the first address past the end of the text segment (the program code). > > + */ > > + > > +extern char etext; > > > > /* > > * int brk(void *addr); > > @@ -554,7 +560,47 @@ long getpagesize(void) > > static __attribute__((unused)) > > int sys_gettimeofday(struct timeval *tv, struct timezone *tz) > > { > > +#ifdef __NR_gettimeofday > > return my_syscall2(__NR_gettimeofday, tv, tz); > > +#elif defined(__NR_clock_gettime) || defined(__NR_clock_gettime64) > > +#ifdef __NR_clock_gettime > > + struct timespec ts; > > +#else > > + struct timespec64 ts; > > +#define __NR_clock_gettime __NR_clock_gettime64 > > +#endif > > + int ret; > > + > > + /* make sure tv pointer is at least after code segment */ > > + if (tv != NULL && (char *)tv <= &etext) > > + return -EFAULT; > > To me the weird etext comparisions don't seem to be worth it, to be > honest. > This is the issue we explained in commit message: * Both tv and tz are not directly passed to kernel clock_gettime* syscalls, so, it isn't able to check the pointer automatically with the get_user/put_user helpers just like kernel gettimeofday syscall does. instead, we emulate (but not completely) such checks in our new __NR_clock_gettime* branch of nolibc. but not that deeply described the direct cause, the direct cause is that the test case passes a '(void *)1' and the kernel space of gettimeofday can simply 'fixup' this issue by the get_user/put_user helpers, but our user-space tv and tz code has no such function, just emulate such 'fixup' by a stupid etext compare to at least make sure the data pointer is in data range. Welcome better solution. CASE_TEST(gettimeofday_bad1); EXPECT_SYSER(1, gettimeofday((void *)1, NULL), -1, EFAULT); break; CASE_TEST(gettimeofday_bad2); EXPECT_SYSER(1, gettimeofday(NULL, (void *)1), -1, EFAULT); break; Without this ugly check, the above cases would get such error: 35 gettimeofday_bad1init[1]: unhandled signal 11 code 0x1 at 0x00000002 in init[10000+5000] CPU: 0 PID: 1 Comm: init Not tainted 6.4.0-rc1-00134-gf929c7b7184f-dirty #20 Hardware name: riscv-virtio,qemu (DT) epc : 00012ccc ra : 00012ca8 sp : 9d254d90 gp : 00016800 tp : 00000000 t0 : 00000000 t1 : 0000000a t2 : 00000000 s0 : 00000001 s1 : 00016008 a0 : 00000000 a1 : 9d254da8 a2 : 00000014 a3 : 00000000 a4 : 00000000 a5 : 00000000 a6 : 00000001 a7 : 00000193 s2 : 00000023 s3 : 9d254da4 s4 : 00000000 s5 : 00000000 s6 : 0000541b s7 : 00000007 s8 : 9d254dcc s9 : 000144e8 s10: 00016000 s11: 00000006 t3 : 00000000 t4 : ffffffff t5 : 00000000 t6 : 00000000 status: 00000020 badaddr: 00000002 cause: 0000000f Will at least append this test error in the commit message of the coming new revision of this patch. Hi, Willy, this also require your discussion, simply remove the above two test cases may be not a good idea too, the check for gettimeofday is perfectly ok. The same 'emulate' method is used in the waitid patch, but that only requires to compare 'pid == INT_MIN', which is not that weird. > > + > > + /* set tz to zero to avoid random number */ > > + if (tz != NULL) { > > + if ((char *)tz > &etext) > > + memset(tz, 0, sizeof(struct timezone)); > > + else > > + return -EFAULT; > > + } > > + The same issue here. > > + if (tv == NULL) > > + return 0; > > + > > + ret = my_syscall2(__NR_clock_gettime, CLOCK_REALTIME, &ts); > > + if (ret) > > + return ret; > > + > > + tv->tv_sec = (time_t) ts.tv_sec; > > +#ifdef __NR_clock_gettime64 > > Nitpick: > > While this ifdef works and is correct its logic is a bit indirect. > If it is copied to some other function in the future it may be incorrect > there. > > Without the #ifdef the compiler should still be able to optimize the > code away. Ok, will remove the #ifdef wrapper, let the compiler optimize itself. Thanks, Zhangjin > > > + if (tv->tv_sec != ts.tv_sec) > > + return -EOVERFLOW; > > +#endif > > + > > + tv->tv_usec = ts.tv_nsec / 1000; > > + return 0; > > +#else > > +#error None of __NR_gettimeofday, __NR_clock_gettime and __NR_clock_gettime64 defined, cannot implement sys_gettimeofday() > > +#endif > > } > > > > static __attribute__((unused)) > > -- > > 2.25.1 > >
WARNING: multiple messages have this Message-ID (diff)
From: Zhangjin Wu <falcon@tinylab.org> To: thomas@t-8ch.de, w@1wt.eu Cc: falcon@tinylab.org, linux-kernel@vger.kernel.org, linux-kselftest@vger.kernel.org, linux-riscv@lists.infradead.org, palmer@dabbelt.com, paul.walmsley@sifive.com Subject: Re: [PATCH 13/13] tools/nolibc: sys_gettimeofday: riscv: use __NR_clock_gettime64 for rv32 Date: Sat, 27 May 2023 09:26:35 +0800 [thread overview] Message-ID: <20230527012635.19595-1-falcon@tinylab.org> (raw) In-Reply-To: <29e4b23d-27cb-4101-bfe9-c6b412222578@t-8ch.de> Hi, Thomas, Willy > On 2023-05-25 02:03:32+0800, Zhangjin Wu wrote: > > rv32 uses the generic include/uapi/asm-generic/unistd.h and it has no > > __NR_gettimeofday and __NR_clock_gettime after kernel commit d4c08b9776b3 > > ("riscv: Use latest system call ABI"), use __NR_clock_gettime64 instead. > > > > This code is based on src/time/gettimeofday.c of musl and > > sysdeps/unix/sysv/linux/clock_gettime.c of glibc. > > > > Both __NR_clock_gettime and __NR_clock_gettime64 are added for > > sys_gettimeofday() for they share most of the code. > > > > Notes: > > > > * Both tv and tz are not directly passed to kernel clock_gettime* > > syscalls, so, it isn't able to check the pointer automatically with the > > get_user/put_user helpers just like kernel gettimeofday syscall does. > > instead, we emulate (but not completely) such checks in our new > > __NR_clock_gettime* branch of nolibc. > > > > * kernel clock_gettime* syscalls can not get tz info, just like musl and > > glibc do, we set tz to zero to avoid a random number. > > > > Signed-off-by: Zhangjin Wu <falcon@tinylab.org> > > --- > > tools/include/nolibc/sys.h | 46 ++++++++++++++++++++++++++++++++++++++ > > 1 file changed, 46 insertions(+) > > > > diff --git a/tools/include/nolibc/sys.h b/tools/include/nolibc/sys.h > > index 2642b380c6aa..ad38cc3856be 100644 > > --- a/tools/include/nolibc/sys.h > > +++ b/tools/include/nolibc/sys.h > > @@ -26,6 +26,7 @@ > > > > #include "arch.h" > > #include "errno.h" > > +#include "string.h" > > #include "types.h" > > > > > > @@ -51,6 +52,11 @@ > > * should not be placed here. > > */ > > > > +/* > > + * This is the first address past the end of the text segment (the program code). > > + */ > > + > > +extern char etext; > > > > /* > > * int brk(void *addr); > > @@ -554,7 +560,47 @@ long getpagesize(void) > > static __attribute__((unused)) > > int sys_gettimeofday(struct timeval *tv, struct timezone *tz) > > { > > +#ifdef __NR_gettimeofday > > return my_syscall2(__NR_gettimeofday, tv, tz); > > +#elif defined(__NR_clock_gettime) || defined(__NR_clock_gettime64) > > +#ifdef __NR_clock_gettime > > + struct timespec ts; > > +#else > > + struct timespec64 ts; > > +#define __NR_clock_gettime __NR_clock_gettime64 > > +#endif > > + int ret; > > + > > + /* make sure tv pointer is at least after code segment */ > > + if (tv != NULL && (char *)tv <= &etext) > > + return -EFAULT; > > To me the weird etext comparisions don't seem to be worth it, to be > honest. > This is the issue we explained in commit message: * Both tv and tz are not directly passed to kernel clock_gettime* syscalls, so, it isn't able to check the pointer automatically with the get_user/put_user helpers just like kernel gettimeofday syscall does. instead, we emulate (but not completely) such checks in our new __NR_clock_gettime* branch of nolibc. but not that deeply described the direct cause, the direct cause is that the test case passes a '(void *)1' and the kernel space of gettimeofday can simply 'fixup' this issue by the get_user/put_user helpers, but our user-space tv and tz code has no such function, just emulate such 'fixup' by a stupid etext compare to at least make sure the data pointer is in data range. Welcome better solution. CASE_TEST(gettimeofday_bad1); EXPECT_SYSER(1, gettimeofday((void *)1, NULL), -1, EFAULT); break; CASE_TEST(gettimeofday_bad2); EXPECT_SYSER(1, gettimeofday(NULL, (void *)1), -1, EFAULT); break; Without this ugly check, the above cases would get such error: 35 gettimeofday_bad1init[1]: unhandled signal 11 code 0x1 at 0x00000002 in init[10000+5000] CPU: 0 PID: 1 Comm: init Not tainted 6.4.0-rc1-00134-gf929c7b7184f-dirty #20 Hardware name: riscv-virtio,qemu (DT) epc : 00012ccc ra : 00012ca8 sp : 9d254d90 gp : 00016800 tp : 00000000 t0 : 00000000 t1 : 0000000a t2 : 00000000 s0 : 00000001 s1 : 00016008 a0 : 00000000 a1 : 9d254da8 a2 : 00000014 a3 : 00000000 a4 : 00000000 a5 : 00000000 a6 : 00000001 a7 : 00000193 s2 : 00000023 s3 : 9d254da4 s4 : 00000000 s5 : 00000000 s6 : 0000541b s7 : 00000007 s8 : 9d254dcc s9 : 000144e8 s10: 00016000 s11: 00000006 t3 : 00000000 t4 : ffffffff t5 : 00000000 t6 : 00000000 status: 00000020 badaddr: 00000002 cause: 0000000f Will at least append this test error in the commit message of the coming new revision of this patch. Hi, Willy, this also require your discussion, simply remove the above two test cases may be not a good idea too, the check for gettimeofday is perfectly ok. The same 'emulate' method is used in the waitid patch, but that only requires to compare 'pid == INT_MIN', which is not that weird. > > + > > + /* set tz to zero to avoid random number */ > > + if (tz != NULL) { > > + if ((char *)tz > &etext) > > + memset(tz, 0, sizeof(struct timezone)); > > + else > > + return -EFAULT; > > + } > > + The same issue here. > > + if (tv == NULL) > > + return 0; > > + > > + ret = my_syscall2(__NR_clock_gettime, CLOCK_REALTIME, &ts); > > + if (ret) > > + return ret; > > + > > + tv->tv_sec = (time_t) ts.tv_sec; > > +#ifdef __NR_clock_gettime64 > > Nitpick: > > While this ifdef works and is correct its logic is a bit indirect. > If it is copied to some other function in the future it may be incorrect > there. > > Without the #ifdef the compiler should still be able to optimize the > code away. Ok, will remove the #ifdef wrapper, let the compiler optimize itself. Thanks, Zhangjin > > > + if (tv->tv_sec != ts.tv_sec) > > + return -EOVERFLOW; > > +#endif > > + > > + tv->tv_usec = ts.tv_nsec / 1000; > > + return 0; > > +#else > > +#error None of __NR_gettimeofday, __NR_clock_gettime and __NR_clock_gettime64 defined, cannot implement sys_gettimeofday() > > +#endif > > } > > > > static __attribute__((unused)) > > -- > > 2.25.1 > > _______________________________________________ linux-riscv mailing list linux-riscv@lists.infradead.org http://lists.infradead.org/mailman/listinfo/linux-riscv
next prev parent reply other threads:[~2023-05-27 1:26 UTC|newest] Thread overview: 119+ messages / expand[flat|nested] mbox.gz Atom feed top 2023-05-24 17:33 [PATCH 00/13] tools/nolibc: riscv: Add full rv32 support Zhangjin Wu 2023-05-24 17:33 ` Zhangjin Wu 2023-05-24 17:41 ` [PATCH 01/13] Revert "tools/nolibc: riscv: Support __NR_llseek for rv32" Zhangjin Wu 2023-05-24 17:41 ` Zhangjin Wu 2023-05-24 17:44 ` [PATCH 02/13] Revert "selftests/nolibc: Fix up compile error " Zhangjin Wu 2023-05-24 17:44 ` Zhangjin Wu 2023-05-24 17:46 ` [PATCH 03/13] selftests/nolibc: print name instead of number for EOVERFLOW Zhangjin Wu 2023-05-24 18:28 ` Zhangjin Wu 2023-05-24 17:46 ` Zhangjin Wu 2023-05-24 20:23 ` Thomas Weißschuh 2023-05-24 20:23 ` Thomas Weißschuh 2023-05-24 17:48 ` [PATCH 04/13] selftests/nolibc: syscall_args: use __NR_statx for rv32 Zhangjin Wu 2023-05-24 17:48 ` Zhangjin Wu 2023-05-24 19:49 ` Thomas Weißschuh 2023-05-24 19:49 ` Thomas Weißschuh 2023-05-25 7:20 ` Zhangjin Wu 2023-05-25 7:20 ` Zhangjin Wu 2023-05-26 9:21 ` Arnd Bergmann 2023-05-26 9:21 ` Arnd Bergmann 2023-05-26 10:06 ` Willy Tarreau 2023-05-26 10:06 ` Willy Tarreau 2023-05-27 0:58 ` Zhangjin Wu 2023-05-27 0:58 ` Zhangjin Wu 2023-05-24 17:50 ` [PATCH 05/13] selftests/nolibc: riscv: customize makefile " Zhangjin Wu 2023-05-24 17:50 ` Zhangjin Wu 2023-05-26 6:57 ` Thomas Weißschuh 2023-05-26 6:57 ` Thomas Weißschuh 2023-05-26 9:20 ` Zhangjin Wu 2023-05-26 9:20 ` Zhangjin Wu 2023-05-24 17:52 ` [PATCH 06/13] selftests/nolibc: allow specify a bios for qemu Zhangjin Wu 2023-05-24 17:52 ` Zhangjin Wu 2023-05-26 7:00 ` Thomas Weißschuh 2023-05-26 7:00 ` Thomas Weißschuh 2023-05-26 10:25 ` Zhangjin Wu 2023-05-26 10:25 ` Zhangjin Wu 2023-05-26 10:36 ` Conor Dooley 2023-05-26 10:36 ` Conor Dooley 2023-05-26 13:38 ` Zhangjin Wu 2023-05-26 13:38 ` Zhangjin Wu 2023-05-26 15:08 ` Conor Dooley 2023-05-26 15:08 ` Conor Dooley 2023-05-28 7:52 ` Willy Tarreau 2023-05-28 7:52 ` Willy Tarreau 2023-05-24 17:54 ` [PATCH 07/13] selftests/nolibc: remove the duplicated gettimeofday_bad2 Zhangjin Wu 2023-05-24 17:54 ` Zhangjin Wu 2023-05-24 17:55 ` [PATCH 08/13] tools/nolibc: sys_lseek: riscv: use __NR_llseek for rv32 Zhangjin Wu 2023-05-24 17:55 ` Zhangjin Wu 2023-05-24 17:57 ` [PATCH 09/13] tools/nolibc: sys_poll: riscv: use __NR_ppoll_time64 " Zhangjin Wu 2023-05-24 17:57 ` Zhangjin Wu 2023-05-26 7:15 ` Thomas Weißschuh 2023-05-26 7:15 ` Thomas Weißschuh 2023-05-26 9:34 ` Arnd Bergmann 2023-05-26 9:34 ` Arnd Bergmann 2023-05-28 8:25 ` Zhangjin Wu 2023-05-28 8:25 ` Zhangjin Wu 2023-05-28 8:48 ` Arnd Bergmann 2023-05-28 8:48 ` Arnd Bergmann 2023-05-28 10:29 ` Willy Tarreau 2023-05-28 10:29 ` Willy Tarreau 2023-05-28 10:55 ` Arnd Bergmann 2023-05-28 10:55 ` Arnd Bergmann 2023-05-28 11:03 ` Willy Tarreau 2023-05-28 11:03 ` Willy Tarreau 2023-05-24 17:58 ` [PATCH 10/13] tools/nolibc: ppoll/ppoll_time64: add a missing argument Zhangjin Wu 2023-05-24 17:58 ` Zhangjin Wu 2023-05-24 17:59 ` [PATCH 11/13] tools/nolibc: sys_select: riscv: use __NR_pselect6_time64 for rv32 Zhangjin Wu 2023-05-24 17:59 ` Zhangjin Wu 2023-05-24 20:22 ` Thomas Weißschuh 2023-05-24 20:22 ` Thomas Weißschuh 2023-05-25 7:10 ` Zhangjin Wu 2023-05-25 7:10 ` Zhangjin Wu 2023-05-25 7:22 ` Thomas Weißschuh 2023-05-25 7:22 ` Thomas Weißschuh 2023-05-26 1:50 ` Zhangjin Wu 2023-05-26 1:50 ` Zhangjin Wu 2023-05-26 9:19 ` Arnd Bergmann 2023-05-26 9:19 ` Arnd Bergmann 2023-05-26 11:00 ` [PATCH 00/13] tools/nolibc: riscv: Add full rv32 support Zhangjin Wu 2023-05-26 11:00 ` Zhangjin Wu 2023-05-26 11:13 ` Arnd Bergmann 2023-05-26 11:13 ` Arnd Bergmann 2023-05-24 18:02 ` [PATCH 12/13] tools/nolibc: sys_wait4: riscv: use __NR_waitid for rv32 Zhangjin Wu 2023-05-24 18:02 ` Zhangjin Wu 2023-05-24 18:03 ` [PATCH 13/13] tools/nolibc: sys_gettimeofday: riscv: use __NR_clock_gettime64 " Zhangjin Wu 2023-05-24 18:03 ` Zhangjin Wu 2023-05-26 7:38 ` Thomas Weißschuh 2023-05-26 7:38 ` Thomas Weißschuh 2023-05-27 1:26 ` Zhangjin Wu [this message] 2023-05-27 1:26 ` Zhangjin Wu 2023-05-27 3:39 ` Zhangjin Wu 2023-05-27 3:39 ` Zhangjin Wu 2023-05-27 5:12 ` Willy Tarreau 2023-05-27 5:12 ` Willy Tarreau 2023-05-24 18:24 ` [PATCH 00/13] tools/nolibc: riscv: Add full rv32 support Zhangjin Wu 2023-05-24 18:24 ` Zhangjin Wu 2023-05-28 7:59 ` Willy Tarreau 2023-05-28 7:59 ` Willy Tarreau 2023-05-28 8:42 ` Thomas Weißschuh 2023-05-28 8:42 ` Thomas Weißschuh 2023-05-28 9:41 ` Thomas Weißschuh 2023-05-28 9:41 ` Thomas Weißschuh 2023-05-28 10:17 ` Willy Tarreau 2023-05-28 10:17 ` Willy Tarreau 2023-05-28 10:39 ` Zhangjin Wu 2023-05-28 10:39 ` Zhangjin Wu 2023-05-28 11:33 ` Willy Tarreau 2023-05-28 11:33 ` Willy Tarreau 2023-05-28 12:52 ` Zhangjin Wu 2023-05-28 12:52 ` Zhangjin Wu 2023-05-28 13:45 ` Thomas Weißschuh 2023-05-28 13:45 ` Thomas Weißschuh 2023-05-28 18:39 ` Zhangjin Wu 2023-05-28 18:39 ` Zhangjin Wu 2023-05-29 8:45 ` Thomas Weißschuh 2023-05-29 8:45 ` Thomas Weißschuh 2023-05-29 11:31 ` Willy Tarreau 2023-05-29 11:31 ` Willy Tarreau 2023-05-30 10:06 ` Zhangjin Wu 2023-05-30 10:06 ` Zhangjin Wu
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=20230527012635.19595-1-falcon@tinylab.org \ --to=falcon@tinylab.org \ --cc=linux-kernel@vger.kernel.org \ --cc=linux-kselftest@vger.kernel.org \ --cc=linux-riscv@lists.infradead.org \ --cc=palmer@dabbelt.com \ --cc=paul.walmsley@sifive.com \ --cc=thomas@t-8ch.de \ --cc=w@1wt.eu \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: linkBe sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.