Re: svn commit: r313006 - in head: sys/conf sys/libkern sys/libkern/x86 sys/sys tests/sys/kern
On Mon, 30 Jan 2017, Conrad Meyer wrote:
On Mon, Jan 30, 2017 at 9:26 PM, Bruce Evans wrote:
On Tue, 31 Jan 2017, Conrad E. Meyer wrote:
Log:
calculate_crc32c: Add SSE4.2 implementation on x86
This breaks building with gcc-4.2.1,
gcc-4.2.1 is an ancient compiler. Good riddance.
I prefer it.
Added: head/sys/libkern/x86/crc32_sse42.c
==
--- /dev/null 00:00:00 1970 (empty, because file is newly added)
+++ head/sys/libkern/x86/crc32_sse42.c Tue Jan 31 03:26:32 2017
(r313006)
+
+#include
...
Inline asm is much less unportable than intrinsics. kib used the correct
method of .byte's in asms to avoid depending on assembler support for newer
instructions. .byte is still used for clflush on amd64 and i386. It
used to be used for invpcid on amd64. I can't find where it is or was
used for xsave stuff.
Konstantin predicted this complaint in code review (phabricator).
Unfortunately, Clang does not automatically unroll asms, even with the
correct mnemonic. Unrolling is essential to performance below the
by-3 block size (768 bytes in this implementation). Hand unrolling in
C seems to generate less efficient assembly than the compiler's
unrolling.
Unorolling is almost completely useless if the instruction takes 3 cycles
like you said it does. The loop calculations run much faster than that,
so they can run in parallel. However, clang generates horrible code for
inline asms instead of intrinsics.
Simple asms can be compiled even by gcc (apparently old binutils supports
SSE4):
X Index: crc32_sse42.c
X ===
X --- crc32_sse42.c (revision 313007)
X +++ crc32_sse42.c (working copy)
X @@ -38,8 +38,29 @@
X #include
X #endif
X
X -#include
X +static __inline uint32_t
X +_mm_crc32_u8(uint32_t x, uint8_t y)
X +{
X + __asm("crc32b %1,%0" : "+r" (x) : "rm" (y));
X + return (x);
X +}
X
X +static __inline uint32_t __unused
X +_mm_crc32_u32(uint32_t x, int32_t y)
X +{
X + __asm("crc32l %1,%0" : "+r" (x) : "rm" (y));
X + return (x);
X +}
X +
X +#ifdef __amd64__
X +static __inline uint64_t
X +_mm_crc32_u64(uint64_t x, int64_t y)
X +{
X + __asm("crc32q %1,%0" : "+r" (x) : "rm" (y));
X + return (x);
X +}
X +#endif
X +
X /* CRC-32C (iSCSI) polynomial in reversed bit order. */
X #define POLY 0x82f63b78
I only tested this at compile time. clang generates horrible code.
Starting with y in an array in memory, it always copies y to the
stack and does a memory access there. There is nothing special about
these inline asms, so I think clang does the same pessimizations for
most asms in cpufunc.h. jkim got annoyed by this for something like
rdtsc() and changed too much to reduce the problem for just one case.
The differences in the generated code look like:
intrinsincs (old <) vs inline asm (new >):
28,29c29,33
This Inner Loop Header: Depth=1
svn commit: r313010 - in vendor-crypto/openssh/dist: . contrib contrib/cygwin contrib/redhat contrib/suse openbsd-compat openbsd-compat/regress regress regress/misc regress/misc/kexfuzz regress/uni...
Author: des Date: Tue Jan 31 12:29:48 2017 New Revision: 313010 URL: https://svnweb.freebsd.org/changeset/base/313010 Log: Vendor import of OpenSSH 7.3p1. Added: vendor-crypto/openssh/dist/.skipped-commit-ids vendor-crypto/openssh/dist/openbsd-compat/bsd-err.c vendor-crypto/openssh/dist/platform-tracing.c vendor-crypto/openssh/dist/regress/cfginclude.sh vendor-crypto/openssh/dist/regress/misc/ vendor-crypto/openssh/dist/regress/misc/Makefile vendor-crypto/openssh/dist/regress/misc/kexfuzz/ vendor-crypto/openssh/dist/regress/misc/kexfuzz/Makefile vendor-crypto/openssh/dist/regress/misc/kexfuzz/README vendor-crypto/openssh/dist/regress/misc/kexfuzz/kexfuzz.c vendor-crypto/openssh/dist/regress/sshcfgparse.sh vendor-crypto/openssh/dist/regress/unittests/utf8/ vendor-crypto/openssh/dist/regress/unittests/utf8/Makefile vendor-crypto/openssh/dist/regress/unittests/utf8/tests.c vendor-crypto/openssh/dist/utf8.c vendor-crypto/openssh/dist/utf8.h Deleted: vendor-crypto/openssh/dist/.cvsignore vendor-crypto/openssh/dist/openbsd-compat/.cvsignore vendor-crypto/openssh/dist/openbsd-compat/regress/.cvsignore vendor-crypto/openssh/dist/regress/.cvsignore vendor-crypto/openssh/dist/roaming.h vendor-crypto/openssh/dist/scard/ Modified: vendor-crypto/openssh/dist/ChangeLog vendor-crypto/openssh/dist/INSTALL vendor-crypto/openssh/dist/Makefile.in vendor-crypto/openssh/dist/PROTOCOL vendor-crypto/openssh/dist/PROTOCOL.agent vendor-crypto/openssh/dist/PROTOCOL.certkeys vendor-crypto/openssh/dist/PROTOCOL.chacha20poly1305 vendor-crypto/openssh/dist/README vendor-crypto/openssh/dist/audit-linux.c vendor-crypto/openssh/dist/auth-krb5.c vendor-crypto/openssh/dist/auth-options.c vendor-crypto/openssh/dist/auth-pam.c vendor-crypto/openssh/dist/auth-pam.h vendor-crypto/openssh/dist/auth-passwd.c vendor-crypto/openssh/dist/auth-rh-rsa.c vendor-crypto/openssh/dist/auth-rhosts.c vendor-crypto/openssh/dist/auth.c vendor-crypto/openssh/dist/auth.h vendor-crypto/openssh/dist/auth2-chall.c vendor-crypto/openssh/dist/auth2-hostbased.c vendor-crypto/openssh/dist/auth2.c vendor-crypto/openssh/dist/authfile.c vendor-crypto/openssh/dist/canohost.c vendor-crypto/openssh/dist/canohost.h vendor-crypto/openssh/dist/channels.c vendor-crypto/openssh/dist/cipher-bf1.c vendor-crypto/openssh/dist/cipher.c vendor-crypto/openssh/dist/clientloop.c vendor-crypto/openssh/dist/compat.c vendor-crypto/openssh/dist/config.h.in vendor-crypto/openssh/dist/configure vendor-crypto/openssh/dist/configure.ac vendor-crypto/openssh/dist/contrib/cygwin/README vendor-crypto/openssh/dist/contrib/redhat/openssh.spec vendor-crypto/openssh/dist/contrib/ssh-copy-id vendor-crypto/openssh/dist/contrib/suse/openssh.spec vendor-crypto/openssh/dist/defines.h vendor-crypto/openssh/dist/dh.c vendor-crypto/openssh/dist/dh.h vendor-crypto/openssh/dist/kex.c vendor-crypto/openssh/dist/kex.h vendor-crypto/openssh/dist/kexc25519.c vendor-crypto/openssh/dist/kexdh.c vendor-crypto/openssh/dist/kexdhc.c vendor-crypto/openssh/dist/kexdhs.c vendor-crypto/openssh/dist/kexgexs.c vendor-crypto/openssh/dist/key.c vendor-crypto/openssh/dist/log.c vendor-crypto/openssh/dist/log.h vendor-crypto/openssh/dist/mac.c vendor-crypto/openssh/dist/mac.h vendor-crypto/openssh/dist/misc.c vendor-crypto/openssh/dist/misc.h vendor-crypto/openssh/dist/moduli vendor-crypto/openssh/dist/moduli.0 vendor-crypto/openssh/dist/monitor.c vendor-crypto/openssh/dist/monitor_fdpass.c vendor-crypto/openssh/dist/monitor_wrap.c vendor-crypto/openssh/dist/monitor_wrap.h vendor-crypto/openssh/dist/mux.c vendor-crypto/openssh/dist/myproposal.h vendor-crypto/openssh/dist/opacket.h vendor-crypto/openssh/dist/openbsd-compat/Makefile.in vendor-crypto/openssh/dist/openbsd-compat/arc4random.c vendor-crypto/openssh/dist/openbsd-compat/bindresvport.c vendor-crypto/openssh/dist/openbsd-compat/bsd-asprintf.c vendor-crypto/openssh/dist/openbsd-compat/bsd-misc.c vendor-crypto/openssh/dist/openbsd-compat/bsd-misc.h vendor-crypto/openssh/dist/openbsd-compat/bsd-snprintf.c vendor-crypto/openssh/dist/openbsd-compat/inet_aton.c vendor-crypto/openssh/dist/openbsd-compat/openbsd-compat.h vendor-crypto/openssh/dist/openbsd-compat/port-solaris.h vendor-crypto/openssh/dist/openbsd-compat/vis.c vendor-crypto/openssh/dist/openbsd-compat/vis.h vendor-crypto/openssh/dist/openbsd-compat/xcrypt.c vendor-crypto/openssh/dist/packet.c vendor-crypto/openssh/dist/packet.h vendor-crypto/openssh/dist/pathnames.h vendor-crypto/openssh/dist/platform.c vendor-crypto/openssh/dist/platform.h vendor-crypto/openssh/dist/progressmeter.c vendor-crypto/openssh/dist/readconf.c vendor-crypto/openssh/dist/readconf.h vendor-crypto/openssh/dist/regress/Makefile vendor-crypto/openssh/dist/regress/agent-getpeereid.sh vendor-crypto/openssh/dist/regress/cert-hostkey.s
svn commit: r313011 - vendor-crypto/openssh/7.3p1
Author: des Date: Tue Jan 31 12:31:38 2017 New Revision: 313011 URL: https://svnweb.freebsd.org/changeset/base/313011 Log: Tag OpenSSH 7.3p1. Added: vendor-crypto/openssh/7.3p1/ - copied from r313010, vendor-crypto/openssh/dist/ ___ svn-src-all@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r313012 - in vendor-crypto/openssh/dist: . contrib contrib/redhat contrib/suse openbsd-compat regress regress/misc/kexfuzz regress/unittests regress/unittests/bitmap regress/unittests/h...
Author: des Date: Tue Jan 31 12:33:47 2017 New Revision: 313012 URL: https://svnweb.freebsd.org/changeset/base/313012 Log: Vendor import of OpenSSH 7.4p1. Added: vendor-crypto/openssh/dist/openbsd-compat/strcasestr.c vendor-crypto/openssh/dist/regress/allow-deny-users.sh vendor-crypto/openssh/dist/regress/keygen-moduli.sh vendor-crypto/openssh/dist/regress/moduli.in vendor-crypto/openssh/dist/regress/unittests/match/ vendor-crypto/openssh/dist/regress/unittests/match/Makefile vendor-crypto/openssh/dist/regress/unittests/match/tests.c Deleted: vendor-crypto/openssh/dist/auth-chall.c vendor-crypto/openssh/dist/auth-rh-rsa.c vendor-crypto/openssh/dist/auth-rsa.c vendor-crypto/openssh/dist/auth1.c vendor-crypto/openssh/dist/monitor_mm.c vendor-crypto/openssh/dist/monitor_mm.h vendor-crypto/openssh/dist/openbsd-compat/xmmap.c Modified: vendor-crypto/openssh/dist/.skipped-commit-ids vendor-crypto/openssh/dist/CREDITS vendor-crypto/openssh/dist/ChangeLog vendor-crypto/openssh/dist/INSTALL vendor-crypto/openssh/dist/Makefile.in vendor-crypto/openssh/dist/PROTOCOL vendor-crypto/openssh/dist/README vendor-crypto/openssh/dist/README.platform vendor-crypto/openssh/dist/README.privsep vendor-crypto/openssh/dist/TODO vendor-crypto/openssh/dist/aclocal.m4 vendor-crypto/openssh/dist/addrmatch.c vendor-crypto/openssh/dist/atomicio.c vendor-crypto/openssh/dist/audit-bsm.c vendor-crypto/openssh/dist/audit-linux.c vendor-crypto/openssh/dist/audit.c vendor-crypto/openssh/dist/audit.h vendor-crypto/openssh/dist/auth-options.c vendor-crypto/openssh/dist/auth-options.h vendor-crypto/openssh/dist/auth-pam.c vendor-crypto/openssh/dist/auth-pam.h vendor-crypto/openssh/dist/auth-rhosts.c vendor-crypto/openssh/dist/auth.c vendor-crypto/openssh/dist/auth.h vendor-crypto/openssh/dist/auth2-pubkey.c vendor-crypto/openssh/dist/authfile.c vendor-crypto/openssh/dist/buildpkg.sh.in vendor-crypto/openssh/dist/chacha.h vendor-crypto/openssh/dist/channels.c vendor-crypto/openssh/dist/channels.h vendor-crypto/openssh/dist/cipher-3des1.c vendor-crypto/openssh/dist/cipher-bf1.c vendor-crypto/openssh/dist/cipher-chachapoly.c vendor-crypto/openssh/dist/cipher.c vendor-crypto/openssh/dist/cipher.h vendor-crypto/openssh/dist/clientloop.c vendor-crypto/openssh/dist/clientloop.h vendor-crypto/openssh/dist/config.guess vendor-crypto/openssh/dist/config.h.in vendor-crypto/openssh/dist/config.sub vendor-crypto/openssh/dist/configure vendor-crypto/openssh/dist/configure.ac vendor-crypto/openssh/dist/contrib/Makefile vendor-crypto/openssh/dist/contrib/gnome-ssh-askpass2.c vendor-crypto/openssh/dist/contrib/redhat/openssh.spec vendor-crypto/openssh/dist/contrib/suse/openssh.spec vendor-crypto/openssh/dist/defines.h vendor-crypto/openssh/dist/dh.c vendor-crypto/openssh/dist/entropy.h vendor-crypto/openssh/dist/gss-genr.c vendor-crypto/openssh/dist/hostfile.c vendor-crypto/openssh/dist/kex.c vendor-crypto/openssh/dist/kex.h vendor-crypto/openssh/dist/kexgexc.c vendor-crypto/openssh/dist/kexgexs.c vendor-crypto/openssh/dist/key.h vendor-crypto/openssh/dist/krl.c vendor-crypto/openssh/dist/mac.c vendor-crypto/openssh/dist/match.c vendor-crypto/openssh/dist/md5crypt.h vendor-crypto/openssh/dist/mdoc2man.awk vendor-crypto/openssh/dist/misc.c vendor-crypto/openssh/dist/misc.h vendor-crypto/openssh/dist/moduli vendor-crypto/openssh/dist/moduli.c vendor-crypto/openssh/dist/monitor.c vendor-crypto/openssh/dist/monitor.h vendor-crypto/openssh/dist/monitor_wrap.c vendor-crypto/openssh/dist/monitor_wrap.h vendor-crypto/openssh/dist/mux.c vendor-crypto/openssh/dist/myproposal.h vendor-crypto/openssh/dist/opacket.h vendor-crypto/openssh/dist/openbsd-compat/Makefile.in vendor-crypto/openssh/dist/openbsd-compat/base64.h vendor-crypto/openssh/dist/openbsd-compat/bsd-asprintf.c vendor-crypto/openssh/dist/openbsd-compat/bsd-cray.c vendor-crypto/openssh/dist/openbsd-compat/bsd-cray.h vendor-crypto/openssh/dist/openbsd-compat/bsd-cygwin_util.c vendor-crypto/openssh/dist/openbsd-compat/bsd-cygwin_util.h vendor-crypto/openssh/dist/openbsd-compat/bsd-misc.c vendor-crypto/openssh/dist/openbsd-compat/bsd-misc.h vendor-crypto/openssh/dist/openbsd-compat/bsd-nextstep.c vendor-crypto/openssh/dist/openbsd-compat/bsd-nextstep.h vendor-crypto/openssh/dist/openbsd-compat/bsd-openpty.c vendor-crypto/openssh/dist/openbsd-compat/bsd-poll.c vendor-crypto/openssh/dist/openbsd-compat/bsd-setres_id.c vendor-crypto/openssh/dist/openbsd-compat/bsd-setres_id.h vendor-crypto/openssh/dist/openbsd-compat/bsd-statvfs.c vendor-crypto/openssh/dist/openbsd-compat/bsd-statvfs.h vendor-crypto/openssh/dist/openbsd-compat/bsd-waitpid.c vendor-crypto/openssh/dist/openbsd-compat/bsd-waitpid.h vendor-crypto/openssh/dist/openbsd-compat/explicit_bzero.c vendor-crypto/openssh/dist/openbsd-compat/fake-rfc2
svn commit: r313013 - vendor-crypto/openssh/7.4p1
Author: des Date: Tue Jan 31 12:33:59 2017 New Revision: 313013 URL: https://svnweb.freebsd.org/changeset/base/313013 Log: Tag OpenSSH 7.4p1. Added: vendor-crypto/openssh/7.4p1/ - copied from r313012, vendor-crypto/openssh/dist/ ___ svn-src-all@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
Re: svn commit: r312702 - in head/sys: kern libkern sys
On Wed, 25 Jan 2017, Bruce Evans wrote: On Tue, 24 Jan 2017, Conrad E. Meyer wrote: Log: Use time_t for intermediate values to avoid overflow in clock_ts_to_ct This is bogus. time_t is for storing times in seconds, not for times in days, hours or minutes. These bugs are still present. Bounds checking for presposterous values in privileged syscalls is silly, especially it is incomplete. There are thousands of sysctls that privileged users can use to shoot themself even more easily. Just near the settimeofday() syscall, similar foot shooting can be accomplished directly using the adjkerntz sysctl. Changes to this normally update the RTC immediately. This is correct if the changes are correct. Small garbage values give garbage in the RTC and large garbage values give overflows. Bounds checking in this commit sometimes detects these overflows. Foot-shooting can also be accomplished indirectly using adjkerntz. Set it to a large (in magnitude) value that doesn't cause overflows. This augments the overflows in the next call to settimeofday(). The fix is not to add bounds checking to all sysctls. Even restricting according to securelevel wouldn't work right, since securelevel doesn't work right. In sys/kern, there are only 10 sysctls (not includinging adjkerntz) restricted by securelevel. However, settimeofday() is restricted by securelevel. You can't set the time backwards at high securelevels, but you can still set the RTC to anything using adjkerntz at high securelevels. settimeofday() gives another unsecured method with results similar to adjkerntz for panicing and corrupting the RTC. set/gettimeofday() still support timezones, and resettodr() adds both adjkerntz and (tz_minuteswest * 60) to convert the kernel time to the RTC time. Both of these additions are FreeBSD features. adjkerntz is necessary to support normal use of the RTC on x86 (it is on local time), but timezones in the kernel were deprecated 25 years ago, so why use them now? On my systems, adjkerntz is always used and I never used tz_minutewest until today to demonstrate foot shooting. The man page says that the timezone is no longer used; actually, it is used for this foot shooting. There are many bugs in the implementation of the foot-shooting: A: There are no bounds check on tz_minuteswest, so expressions like (tz_minuteswest * 60) may overflow before the foot is the target. settimeofday() tries to be careful about the order of operations, but there is no order that just works for the dependent operations of setting the time and the timezone. The new timezone should be set before setting the time, since the timezone is needed for the RTC part of setting the time, and settings should be reversible in case one part fails. In practice, the time is set first, and if this succeeds then further errors are mostly not detected, but sometimes cause panics; then the RTC is set using a non-atomic copy of the pair (adjkerntz, tz_minuteswest), where tz_minuteswest was usually set by the previous settimeofday() call. Then tz_minuteswest is set for this call. B: There is no interlocking for any of this. adjkerntz and tz_minuteswest are plain ints, so reading them is probably atomic, but any number of adjkerntz sysctls and settimeofday() syscalls can race each other. The adjkerntz sysctl claims to be MPSAFE, but does nothing special. It just overwrites the adjkerntz using a hopefully-atomic access, then calls resettodr() like settime(). kib added some locking around the hardware part of resettodr(), and did more careful things for the boottime offset for timecounters. The boottime offset is more important since it is used for every CLOCK_REALTIME timecounter call, and needs atomic accesses more since it is larger than an int. Some of the interlocking belongs in applications. E.g., after setting the time, read the time and the monotonic time to check that the (monotonic) time taken to set the time was not so long as to make the setting too out of date, and retry until it was not too out of date. Coherency between separate settings of the time and the RTC could be maintained in the same way, but since the kernel sets both it do the check. Granularity of the x86 RTC is 1 second, and accuracy is worse, so a long lag is acceptable, but this is MD. My version doesn't bother doing RTC updates for differences of less than 2 seconds. Add additionally safety and overflow checks to clock_ts_to_ct and the BCD routines while we're here. I also disagreed with previous versions of this fix. Perform a safety check in sys_clock_settime() first to avoid easy local root panic, without having to propagate an error value back through dozens of APIs currently lacking error returns. I agree with not over-engineering this to check at all levels. But top-level check needs to be more stringent and magic to work. It is easier to check a couple of levels lower. But there is no way for lower levels to report errors, so the up-fr
svn commit: r313014 - head/sys/boot/fdt/dts/arm
Author: jmcneill Date: Tue Jan 31 14:36:06 2017 New Revision: 313014 URL: https://svnweb.freebsd.org/changeset/base/313014 Log: Add missing ephy bus gates and resets. Modified: head/sys/boot/fdt/dts/arm/h3.dtsi Modified: head/sys/boot/fdt/dts/arm/h3.dtsi == --- head/sys/boot/fdt/dts/arm/h3.dtsi Tue Jan 31 12:33:59 2017 (r313013) +++ head/sys/boot/fdt/dts/arm/h3.dtsi Tue Jan 31 14:36:06 2017 (r313014) @@ -69,10 +69,10 @@ reg = <0x01c3 0x104>, <0x01c00030 0x4>; reg-names = "emac", "syscon"; interrupts = ; - resets = <&ahb_rst 17>; - reset-names = "ahb"; - clocks = <&bus_gates 17>; - clock-names = "ahb"; + resets = <&ahb_rst 17>, <&ahb_rst 66>; + reset-names = "ahb", "ephy"; + clocks = <&bus_gates 17>, <&bus_gates 128>; + clock-names = "ahb", "ephy"; #address-cells = <1>; #size-cells = <0>; status = "disabled"; ___ svn-src-all@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
Re: svn commit: r312988 - in head/sys: compat/cloudabi compat/linux kern sys
On 0130T1502, Gleb Smirnoff wrote: > On Mon, Jan 30, 2017 at 12:57:23PM +, Edward Tomasz Napierala wrote: > E> Author: trasz > E> Date: Mon Jan 30 12:57:22 2017 > E> New Revision: 312988 > E> URL: https://svnweb.freebsd.org/changeset/base/312988 > E> > E> Log: > E> Add kern_listen(), kern_shutdown(), and kern_socket(), and use them > E> instead of their sys_*() counterparts in various compats. The svr4 > E> is left untouched, because there's no point. > > Btw, any good reasons to keep SVR4 support for FreeBSD 12? I don't think so. However, I seem to remember there was an attempt to remove one of the compats some time ago, and it turned out there were still some people using it. So, I don't have plans to disconnect svr4, but I don't have plans to keep it up to date either. ___ svn-src-all@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
Re: svn commit: r313008 - head/tests/sys/geom/class/gate
I think wait_for_ggate_device should also be used at line 32, which
calls "ggatec create".
-Alan
On Mon, Jan 30, 2017 at 11:12 PM, Ngie Cooper wrote:
> Author: ngie
> Date: Tue Jan 31 06:12:51 2017
> New Revision: 313008
> URL: https://svnweb.freebsd.org/changeset/base/313008
>
> Log:
> Wait for /dev/ggate* to appear after calling `ggatel create` in
> :ggatel_{file,md}
>
> The test assumed that `ggatel create` created a device on completion, but
> that's
> incorrect. This squashes the race by waiting for the device to appear, as
> `ggatel create` daemonizes before issuing an ioctl to geom_gate(4) if not
> called
> with `-v`.
>
> Discussed with: asomers
> MFC after:1 week
> PR: 204616
> Sponsored by: Dell EMC Isilon
>
> Modified:
> head/tests/sys/geom/class/gate/ggate_test.sh
>
> Modified: head/tests/sys/geom/class/gate/ggate_test.sh
> ==
> --- head/tests/sys/geom/class/gate/ggate_test.shTue Jan 31 03:40:13
> 2017(r313007)
> +++ head/tests/sys/geom/class/gate/ggate_test.shTue Jan 31 06:12:51
> 2017(r313008)
> @@ -74,7 +74,11 @@ ggatel_file_body()
>
> atf_check ggatel create -u $us work
>
> - dd if=src of=/dev/ggate${us} bs=1m count=1 conv=notrunc
> + ggate_dev=/dev/ggate${us}
> +
> + wait_for_ggate_device ${ggate_dev}
> +
> + dd if=src of=${ggate_dev} bs=1m count=1 conv=notrunc
>
> checksum src work
> }
> @@ -104,7 +108,11 @@ ggatel_md_body()
>
> atf_check ggatel create -u $us /dev/$work
>
> - dd if=/dev/$src of=/dev/ggate${us} bs=1m count=1 conv=notrunc
> + ggate_dev=/dev/ggate${us}
> +
> + wait_for_ggate_device ${ggate_dev}
> +
> + dd if=/dev/$src of=${ggate_dev} bs=1m count=1 conv=notrunc
>
> checksum /dev/$src /dev/$work
> }
> @@ -191,3 +199,14 @@ common_cleanup()
> fi
> true
> }
> +
> +# Bug 204616: ggatel(8) creates /dev/ggate* asynchronously if `ggatel create`
> +# isn't called with `-v`.
> +wait_for_ggate_device()
> +{
> + ggate_device=$1
> +
> + while [ ! -c $ggate_device ]; do
> + sleep 0.5
> + done
> +}
>
___
svn-src-all@freebsd.org mailing list
https://lists.freebsd.org/mailman/listinfo/svn-src-all
To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r313015 - in head/sys: compat/freebsd32 kern sys
Author: trasz
Date: Tue Jan 31 15:11:23 2017
New Revision: 313015
URL: https://svnweb.freebsd.org/changeset/base/313015
Log:
Add kern_cpuset_getid() and kern_cpuset_setid(), and use them
in compat32 instead of their sub_*() counterparts.
Reviewed by: jhb@, kib@
MFC after:2 weeks
Sponsored by: DARPA, AFRL
Differential Revision:https://reviews.freebsd.org/D9382
Modified:
head/sys/compat/freebsd32/freebsd32_misc.c
head/sys/kern/kern_cpuset.c
head/sys/sys/syscallsubr.h
Modified: head/sys/compat/freebsd32/freebsd32_misc.c
==
--- head/sys/compat/freebsd32/freebsd32_misc.c Tue Jan 31 14:36:06 2017
(r313014)
+++ head/sys/compat/freebsd32/freebsd32_misc.c Tue Jan 31 15:11:23 2017
(r313015)
@@ -2556,27 +2556,18 @@ int
freebsd32_cpuset_setid(struct thread *td,
struct freebsd32_cpuset_setid_args *uap)
{
- struct cpuset_setid_args ap;
- ap.which = uap->which;
- ap.id = PAIR32TO64(id_t,uap->id);
- ap.setid = uap->setid;
-
- return (sys_cpuset_setid(td, &ap));
+ return (kern_cpuset_setid(td, uap->which,
+ PAIR32TO64(id_t, uap->id), uap->setid));
}
int
freebsd32_cpuset_getid(struct thread *td,
struct freebsd32_cpuset_getid_args *uap)
{
- struct cpuset_getid_args ap;
-
- ap.level = uap->level;
- ap.which = uap->which;
- ap.id = PAIR32TO64(id_t,uap->id);
- ap.setid = uap->setid;
- return (sys_cpuset_getid(td, &ap));
+ return (kern_cpuset_getid(td, uap->level, uap->which,
+ PAIR32TO64(id_t, uap->id), uap->setid));
}
int
Modified: head/sys/kern/kern_cpuset.c
==
--- head/sys/kern/kern_cpuset.c Tue Jan 31 14:36:06 2017(r313014)
+++ head/sys/kern/kern_cpuset.c Tue Jan 31 15:11:23 2017(r313015)
@@ -976,18 +976,26 @@ struct cpuset_setid_args {
int
sys_cpuset_setid(struct thread *td, struct cpuset_setid_args *uap)
{
+
+ return (kern_cpuset_setid(td, uap->which, uap->id, uap->setid));
+}
+
+int
+kern_cpuset_setid(struct thread *td, cpuwhich_t which,
+id_t id, cpusetid_t setid)
+{
struct cpuset *set;
int error;
/*
* Presently we only support per-process sets.
*/
- if (uap->which != CPU_WHICH_PID)
+ if (which != CPU_WHICH_PID)
return (EINVAL);
- set = cpuset_lookup(uap->setid, td);
+ set = cpuset_lookup(setid, td);
if (set == NULL)
return (ESRCH);
- error = cpuset_setproc(uap->id, set, NULL);
+ error = cpuset_setproc(id, set, NULL);
cpuset_rel(set);
return (error);
}
@@ -1003,19 +1011,28 @@ struct cpuset_getid_args {
int
sys_cpuset_getid(struct thread *td, struct cpuset_getid_args *uap)
{
+
+ return (kern_cpuset_getid(td, uap->level, uap->which, uap->id,
+ uap->setid));
+}
+
+int
+kern_cpuset_getid(struct thread *td, cpulevel_t level, cpuwhich_t which,
+id_t id, cpusetid_t *setid)
+{
struct cpuset *nset;
struct cpuset *set;
struct thread *ttd;
struct proc *p;
- cpusetid_t id;
+ cpusetid_t tmpid;
int error;
- if (uap->level == CPU_LEVEL_WHICH && uap->which != CPU_WHICH_CPUSET)
+ if (level == CPU_LEVEL_WHICH && which != CPU_WHICH_CPUSET)
return (EINVAL);
- error = cpuset_which(uap->which, uap->id, &p, &ttd, &set);
+ error = cpuset_which(which, id, &p, &ttd, &set);
if (error)
return (error);
- switch (uap->which) {
+ switch (which) {
case CPU_WHICH_TID:
case CPU_WHICH_PID:
thread_lock(ttd);
@@ -1030,7 +1047,7 @@ sys_cpuset_getid(struct thread *td, stru
case CPU_WHICH_DOMAIN:
return (EINVAL);
}
- switch (uap->level) {
+ switch (level) {
case CPU_LEVEL_ROOT:
nset = cpuset_refroot(set);
cpuset_rel(set);
@@ -1041,10 +1058,10 @@ sys_cpuset_getid(struct thread *td, stru
case CPU_LEVEL_WHICH:
break;
}
- id = set->cs_id;
+ tmpid = set->cs_id;
cpuset_rel(set);
if (error == 0)
- error = copyout(&id, uap->setid, sizeof(id));
+ error = copyout(&tmpid, setid, sizeof(id));
return (error);
}
Modified: head/sys/sys/syscallsubr.h
==
--- head/sys/sys/syscallsubr.h Tue Jan 31 14:36:06 2017(r313014)
+++ head/sys/sys/syscallsubr.h Tue Jan 31 15:11:23 2017(r313015)
@@ -86,6 +86,10 @@ int kern_clock_settime(struct thread *td
intkern_close(struct thread *td, int fd);
intkern_connectat(struct thread *td, int dirfd, int fd,
struct sockaddr *sa);
+intkern_cpuset_getid(struct thread *td, cpulevel
svn commit: r313016 - in head/sys: compat/freebsd32 kern
Author: trasz
Date: Tue Jan 31 15:19:44 2017
New Revision: 313016
URL: https://svnweb.freebsd.org/changeset/base/313016
Log:
Replace calls to sys_truncate() with kern_truncate().
Reviewed by: kib@
MFC after:2 weeks
Sponsored by: DARPA, AFRL
Differential Revision:https://reviews.freebsd.org/D9371
Modified:
head/sys/compat/freebsd32/freebsd32_misc.c
head/sys/kern/vfs_syscalls.c
Modified: head/sys/compat/freebsd32/freebsd32_misc.c
==
--- head/sys/compat/freebsd32/freebsd32_misc.c Tue Jan 31 15:11:23 2017
(r313015)
+++ head/sys/compat/freebsd32/freebsd32_misc.c Tue Jan 31 15:19:44 2017
(r313016)
@@ -1500,11 +1500,9 @@ freebsd32_lseek(struct thread *td, struc
int
freebsd32_truncate(struct thread *td, struct freebsd32_truncate_args *uap)
{
- struct truncate_args ap;
- ap.path = uap->path;
- ap.length = PAIR32TO64(off_t,uap->length);
- return (sys_truncate(td, &ap));
+ return (kern_truncate(td, uap->path, UIO_USERSPACE,
+ PAIR32TO64(off_t, uap->length)));
}
int
@@ -1600,11 +1598,9 @@ freebsd6_freebsd32_lseek(struct thread *
int
freebsd6_freebsd32_truncate(struct thread *td, struct
freebsd6_freebsd32_truncate_args *uap)
{
- struct truncate_args ap;
- ap.path = uap->path;
- ap.length = PAIR32TO64(off_t,uap->length);
- return (sys_truncate(td, &ap));
+ return (kern_truncate(td, uap->path, UIO_USERSPACE,
+ PAIR32TO64(off_t, uap->length)));
}
int
Modified: head/sys/kern/vfs_syscalls.c
==
--- head/sys/kern/vfs_syscalls.cTue Jan 31 15:11:23 2017
(r313015)
+++ head/sys/kern/vfs_syscalls.cTue Jan 31 15:19:44 2017
(r313016)
@@ -3310,22 +3310,10 @@ struct otruncate_args {
};
#endif
int
-otruncate(td, uap)
- struct thread *td;
- register struct otruncate_args /* {
- char *path;
- long length;
- } */ *uap;
+otruncate(struct thread *td, struct otruncate_args *uap)
{
- struct truncate_args /* {
- char *path;
- int pad;
- off_t length;
- } */ nuap;
-
- nuap.path = uap->path;
- nuap.length = uap->length;
- return (sys_truncate(td, &nuap));
+
+ return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
}
#endif /* COMPAT_43 */
@@ -3334,11 +3322,8 @@ otruncate(td, uap)
int
freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
{
- struct truncate_args ouap;
- ouap.path = uap->path;
- ouap.length = uap->length;
- return (sys_truncate(td, &ouap));
+ return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
}
int
___
svn-src-all@freebsd.org mailing list
https://lists.freebsd.org/mailman/listinfo/svn-src-all
To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r313017 - head/sys/compat/linux
Author: trasz Date: Tue Jan 31 15:22:51 2017 New Revision: 313017 URL: https://svnweb.freebsd.org/changeset/base/313017 Log: Fix linux_getppid() to debug the actual parent, even it was reparented by debugger. Reviewed by: dchagin@ MFC after:2 weeks Sponsored by: DARPA, AFRL Differential Revision:https://reviews.freebsd.org/D9361 Modified: head/sys/compat/linux/linux_misc.c Modified: head/sys/compat/linux/linux_misc.c == --- head/sys/compat/linux/linux_misc.c Tue Jan 31 15:19:44 2017 (r313016) +++ head/sys/compat/linux/linux_misc.c Tue Jan 31 15:22:51 2017 (r313017) @@ -1733,9 +1733,7 @@ linux_getppid(struct thread *td, struct printf(ARGS(getppid, "")); #endif - PROC_LOCK(td->td_proc); - td->td_retval[0] = td->td_proc->p_pptr->p_pid; - PROC_UNLOCK(td->td_proc); + td->td_retval[0] = kern_getppid(td); return (0); } ___ svn-src-all@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r313018 - in head/sys: compat/freebsd32 compat/linux kern sys
Author: trasz
Date: Tue Jan 31 15:35:18 2017
New Revision: 313018
URL: https://svnweb.freebsd.org/changeset/base/313018
Log:
Add kern_pread() and kern_pwrite(), and use it in compats instead
of their sys_*() counterparts. The svr4 is left unchanged.
Reviewed by: kib@
MFC after:2 weeks
Sponsored by: DARPA, AFRL
Differential Revision:https://reviews.freebsd.org/D9379
Modified:
head/sys/compat/freebsd32/freebsd32_misc.c
head/sys/compat/linux/linux_file.c
head/sys/kern/sys_generic.c
head/sys/sys/syscallsubr.h
Modified: head/sys/compat/freebsd32/freebsd32_misc.c
==
--- head/sys/compat/freebsd32/freebsd32_misc.c Tue Jan 31 15:22:51 2017
(r313017)
+++ head/sys/compat/freebsd32/freebsd32_misc.c Tue Jan 31 15:35:18 2017
(r313018)
@@ -1452,25 +1452,17 @@ freebsd4_freebsd32_fhstatfs(struct threa
int
freebsd32_pread(struct thread *td, struct freebsd32_pread_args *uap)
{
- struct pread_args ap;
- ap.fd = uap->fd;
- ap.buf = uap->buf;
- ap.nbyte = uap->nbyte;
- ap.offset = PAIR32TO64(off_t,uap->offset);
- return (sys_pread(td, &ap));
+ return (kern_pread(td, uap->fd, uap->buf, uap->nbyte,
+ PAIR32TO64(off_t, uap->offset)));
}
int
freebsd32_pwrite(struct thread *td, struct freebsd32_pwrite_args *uap)
{
- struct pwrite_args ap;
- ap.fd = uap->fd;
- ap.buf = uap->buf;
- ap.nbyte = uap->nbyte;
- ap.offset = PAIR32TO64(off_t,uap->offset);
- return (sys_pwrite(td, &ap));
+ return (kern_pwrite(td, uap->fd, uap->buf, uap->nbyte,
+ PAIR32TO64(off_t, uap->offset)));
}
#ifdef COMPAT_43
@@ -1559,25 +1551,17 @@ freebsd32_getdirentries(struct thread *t
int
freebsd6_freebsd32_pread(struct thread *td, struct
freebsd6_freebsd32_pread_args *uap)
{
- struct pread_args ap;
- ap.fd = uap->fd;
- ap.buf = uap->buf;
- ap.nbyte = uap->nbyte;
- ap.offset = PAIR32TO64(off_t,uap->offset);
- return (sys_pread(td, &ap));
+ return (kern_pread(td, uap->fd, uap->buf, uap->nbyte,
+ PAIR32TO64(off_t, uap->offset)));
}
int
freebsd6_freebsd32_pwrite(struct thread *td, struct
freebsd6_freebsd32_pwrite_args *uap)
{
- struct pwrite_args ap;
- ap.fd = uap->fd;
- ap.buf = uap->buf;
- ap.nbyte = uap->nbyte;
- ap.offset = PAIR32TO64(off_t,uap->offset);
- return (sys_pwrite(td, &ap));
+ return (kern_pwrite(td, uap->fd, uap->buf, uap->nbyte,
+ PAIR32TO64(off_t, uap->offset)));
}
int
Modified: head/sys/compat/linux/linux_file.c
==
--- head/sys/compat/linux/linux_file.c Tue Jan 31 15:22:51 2017
(r313017)
+++ head/sys/compat/linux/linux_file.c Tue Jan 31 15:35:18 2017
(r313018)
@@ -997,20 +997,13 @@ linux_fdatasync(td, uap)
}
int
-linux_pread(td, uap)
- struct thread *td;
- struct linux_pread_args *uap;
+linux_pread(struct thread *td, struct linux_pread_args *uap)
{
- struct pread_args bsd;
cap_rights_t rights;
struct vnode *vp;
int error;
- bsd.fd = uap->fd;
- bsd.buf = uap->buf;
- bsd.nbyte = uap->nbyte;
- bsd.offset = uap->offset;
- error = sys_pread(td, &bsd);
+ error = kern_pread(td, uap->fd, uap->buf, uap->nbyte, uap->offset);
if (error == 0) {
/* This seems to violate POSIX but linux does it */
error = fgetvp(td, uap->fd,
@@ -1027,17 +1020,10 @@ linux_pread(td, uap)
}
int
-linux_pwrite(td, uap)
- struct thread *td;
- struct linux_pwrite_args *uap;
-{
- struct pwrite_args bsd;
-
- bsd.fd = uap->fd;
- bsd.buf = uap->buf;
- bsd.nbyte = uap->nbyte;
- bsd.offset = uap->offset;
- return (sys_pwrite(td, &bsd));
+linux_pwrite(struct thread *td, struct linux_pwrite_args *uap)
+{
+
+ return (kern_pwrite(td, uap->fd, uap->buf, uap->nbyte, uap->offset));
}
int
Modified: head/sys/kern/sys_generic.c
==
--- head/sys/kern/sys_generic.c Tue Jan 31 15:22:51 2017(r313017)
+++ head/sys/kern/sys_generic.c Tue Jan 31 15:35:18 2017(r313018)
@@ -220,39 +220,37 @@ struct pread_args {
};
#endif
int
-sys_pread(td, uap)
- struct thread *td;
- struct pread_args *uap;
+sys_pread(struct thread *td, struct pread_args *uap)
+{
+
+ return (kern_pread(td, uap->fd, uap->buf, uap->nbyte, uap->offset));
+}
+
+int
+kern_pread(struct thread *td, int fd, void *buf, size_t nbyte, off_t offset)
{
struct uio auio;
struct iovec aiov;
int error;
- if (uap->nbyte > IOSIZE_MAX)
+ if (nbyte > IOSIZE_MAX)
return (EINVAL);
- aiov.iov_base = uap->buf;
- aiov.iov_len = uap->nbyte;
+
Re: svn commit: r313006 - in head: sys/conf sys/libkern sys/libkern/x86 sys/sys tests/sys/kern
On Tue, 31 Jan 2017, Bruce Evans wrote:
On Mon, 30 Jan 2017, Conrad Meyer wrote:
On Mon, Jan 30, 2017 at 9:26 PM, Bruce Evans wrote:
On Tue, 31 Jan 2017, Conrad E. Meyer wrote:
Log:
calculate_crc32c: Add SSE4.2 implementation on x86
This breaks building with gcc-4.2.1,
gcc-4.2.1 is an ancient compiler. Good riddance.
I prefer it.
It also works better on ordinary asms for crc32.
Added: head/sys/libkern/x86/crc32_sse42.c
==
--- /dev/null 00:00:00 1970 (empty, because file is newly added)
+++ head/sys/libkern/x86/crc32_sse42.c Tue Jan 31 03:26:32 2017
(r313006)
+
+#include
...
Inline asm is much less unportable than intrinsics. kib used the correct
method of .byte's in asms to avoid depending on assembler support for
newer
instructions. .byte is still used for clflush on amd64 and i386. It
used to be used for invpcid on amd64. I can't find where it is or was
used for xsave stuff.
Konstantin predicted this complaint in code review (phabricator).
Unfortunately, Clang does not automatically unroll asms, even with the
correct mnemonic. Unrolling is essential to performance below the
by-3 block size (768 bytes in this implementation). Hand unrolling in
C seems to generate less efficient assembly than the compiler's
unrolling.
Unorolling is almost completely useless if the instruction takes 3 cycles
like you said it does. The loop calculations run much faster than that,
so they can run in parallel. However, clang generates horrible code for
inline asms instead of intrinsics.
The unrolling was worse than useless (see previous reply). It is the
horrible code in the non-intrinsics case that seems to be the main source
of differences (see below).
The left column below is block size. The measurements are nanoseconds
per buf, per CLOCK_VIRTUAL, averaged over 10^5 loops. These numbers
do not vary more than +/- 1ns run to run on my idle Sandy Bridge
laptop. "asm" is using __asm__(), "intrins" using the _mm_crc32
intrinsics that Clang can unroll, and multitable is the older
lookup-table implementation (still used on other architectures).
0x10: asm:0 intrins:0 multitable:0 (ns per buf)
0x20: asm:7 intrins:9 multitable:78 (ns per buf)
0x40: asm:10 intrins:7 multitable:50 (ns per buf)
0x80: asm:15 intrins:9 multitable:91 (ns per buf)
0x000100: asm:25 intrins:17 multitable:178 (ns per buf)
0x000200: asm:55 intrins:38 multitable:347 (ns per buf)
0x000400: asm:61 intrins:62 multitable:684 (ns per buf)
Both implementations are superior to the multitable approach, so it is
unreasonable not to make one of them standard on x86 platforms.
The unrolled intrinsics are consistently better than not unrolled on
objects 0x40-0x200 bytes large. At 0x400 bytes we pass the first
unroll-by-3 threshold and it stops mattering as much.
At 0x40 bytes, it is the difference between 6.4 GB/s and 9.1 GB/s. At
0x200 bytes, it is the difference between 9.3 GB/s and 13.5 GB/s. I
think this justifies some minor ugliness.
If it matters, which I doubt, then the compiler shouldn't be trusted
for unrolling. It can only do better than handwritten unrolling if
it has tables for the correct amount unrolling for every instruction
for thousands of CPUs (since such tables are too hard to write manually).
It is the 3-way unrolling that makes the main difference. clang is
clueless about unrolling in this function, but this 3-way unrolling is
in the C code. It is to reduce dependencies. The timing seems to be
as I suspected -- the crc32 instruction has a latency of 3 and a throughput
of 3 per 3 cycles, so with 3 dependent instructions it runs 3 times slower
than with 3 independent instructions. Naive unrolling gives more dependent
instructions, so is equally slow to no unrolling.
Test program for this:
X #include
X #include
X
X #ifndef __clang__
Change this to __clang__X to kill intrinsics for clang.
X static __inline uint32_t
X _mm_crc32_u8(uint32_t x, uint8_t y)
X {
X __asm("crc32b %1,%0" : "+r" (x) : "rm" (y));
X return (x);
X }
X
X static __inline uint32_t __unused
X _mm_crc32_u32(uint32_t x, int32_t y)
X {
X __asm("crc32l %1,%0" : "+r" (x) : "rm" (y));
X return (x);
X }
X
X #ifdef __amd64__
X static __inline uint64_t
X _mm_crc32_u64(uint64_t x, int64_t y)
X {
X __asm("crc32q %1,%0" : "+r" (x) : "rm" (y));
X return (x);
X }
X #endif
X #else
X #include
X #endif
X
X #define MISALIGN 0
MISALIGN 1 costs about 4% (0.36 seconds instead of 0.34 seconds on
freefall's Xeon).
X #define SIZE 100
The large size of 1M is to bust at least the L1 cache. This makes little
difference, since the speed is CPU-bound (0.34 seconds @ 3.3 GHz = ~34/33
cycles per crc32 instruction). Actually, I'm not sure about the speeds.
freefalls's machdep.tsc_freq is 3.3GHz, but the speed could be anything
due to turbo mode and frequency control (boot message give simi
Re: svn commit: r312988 - in head/sys: compat/cloudabi compat/linux kern sys
On Tue, Jan 31, 2017 at 03:07:27PM +, Edward Tomasz Napierala wrote: > On 0130T1502, Gleb Smirnoff wrote: > > On Mon, Jan 30, 2017 at 12:57:23PM +, Edward Tomasz Napierala wrote: > > E> Author: trasz > > E> Date: Mon Jan 30 12:57:22 2017 > > E> New Revision: 312988 > > E> URL: https://svnweb.freebsd.org/changeset/base/312988 > > E> > > E> Log: > > E> Add kern_listen(), kern_shutdown(), and kern_socket(), and use them > > E> instead of their sys_*() counterparts in various compats. The svr4 > > E> is left untouched, because there's no point. > > > > Btw, any good reasons to keep SVR4 support for FreeBSD 12? > > I don't think so. However, I seem to remember there was an attempt > to remove one of the compats some time ago, and it turned out there were > still some people using it. So, I don't have plans to disconnect svr4, > but I don't have plans to keep it up to date either. > SVr4 is barely useful, but still connected to the build. iBCS2, on the other hand, was used by many people: I can claim it because I usually got more than one report for a regression in that code when core parts of the kernel change. I did some improvements for iBCS2 previous summer, both to better fit it into current kernel code, and fix outstanding bugs, biggest of which was incompatible signal frame layout. But there were concerns that the work required an approval from the SCO copyright owners, and we were unable to get a clearing. As result, iBCS2 emulator was disconnected from the build for 11 and HEAD. It would be somewhat pity to kill it because it is useful for many people, but we cannot update the code safely. There is no such concern for SVr4. ___ svn-src-all@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r313019 - in head/sys: dev/e1000 dev/vnic net
Author: stevek
Date: Tue Jan 31 16:12:31 2017
New Revision: 313019
URL: https://svnweb.freebsd.org/changeset/base/313019
Log:
Add the folowing set accessor functions for recently-added members of ifnet
structure:
if_gethwtsomax(), if_sethwtsomax() - if_hw_tsomax
if_gethwtsomaxsegcount(), if_sethwtsomaxsegcount() - if_hw_tsomaxsegcount
if_gethwtsomaxsegsize(), if_sethwtsomaxsegsize() - if_hw_tsomaxsegsize
Update em and vnic drivers which had already been coverted to use accessor
functions for the other ifnet structure members.
Reviewed by: erj
Approved by: sjg (mentor)
Obtained from:Juniper Networks, Inc.
Differential Revision:https://reviews.freebsd.org/D8544
Modified:
head/sys/dev/e1000/if_em.c
head/sys/dev/vnic/nicvf_main.c
head/sys/net/if.c
head/sys/net/if_var.h
Modified: head/sys/dev/e1000/if_em.c
==
--- head/sys/dev/e1000/if_em.c Tue Jan 31 15:35:18 2017(r313018)
+++ head/sys/dev/e1000/if_em.c Tue Jan 31 16:12:31 2017(r313019)
@@ -2481,10 +2481,10 @@ em_setup_interface(if_ctx_t ctx)
INIT_DEBUGOUT("em_setup_interface: begin");
/* TSO parameters */
- ifp->if_hw_tsomax = IP_MAXPACKET;
+ if_sethwtsomax(ifp, IP_MAXPACKET);
/* Take m_pullup(9)'s in em_xmit() w/ TSO into acount. */
- ifp->if_hw_tsomaxsegcount = EM_MAX_SCATTER - 5;
- ifp->if_hw_tsomaxsegsize = EM_TSO_SEG_SIZE;
+ if_sethwtsomaxsegcount(ifp, EM_MAX_SCATTER - 5);
+ if_sethwtsomaxsegsize(ifp, EM_TSO_SEG_SIZE);
/* Single Queue */
if (adapter->tx_num_queues == 1) {
Modified: head/sys/dev/vnic/nicvf_main.c
==
--- head/sys/dev/vnic/nicvf_main.c Tue Jan 31 15:35:18 2017
(r313018)
+++ head/sys/dev/vnic/nicvf_main.c Tue Jan 31 16:12:31 2017
(r313019)
@@ -370,9 +370,9 @@ nicvf_setup_ifnet(struct nicvf *nic)
/* TSO */
if_setcapabilitiesbit(ifp, IFCAP_TSO4, 0);
/* TSO parameters */
- ifp->if_hw_tsomax = NICVF_TSO_MAXSIZE;
- ifp->if_hw_tsomaxsegcount = NICVF_TSO_NSEGS;
- ifp->if_hw_tsomaxsegsize = MCLBYTES;
+ if_sethwtsomax(ifp, NICVF_TSO_MAXSIZE);
+ if_sethwtsomaxsegcount(ifp, NICVF_TSO_NSEGS);
+ if_sethwtsomaxsegsize(ifp, MCLBYTES);
}
/* IP/TCP/UDP HW checksums */
if_setcapabilitiesbit(ifp, IFCAP_HWCSUM, 0);
@@ -453,7 +453,7 @@ nicvf_if_ioctl(struct ifnet *ifp, u_long
#if defined(INET) || defined(INET6)
/* Avoid reinitialization unless it's necessary */
if (avoid_reset) {
- ifp->if_flags |= IFF_UP;
+ if_setflagbits(ifp, IFF_UP, 0);
if (!(if_getdrvflags(ifp) & IFF_DRV_RUNNING))
nicvf_if_init(nic);
#ifdef INET
@@ -482,8 +482,8 @@ nicvf_if_ioctl(struct ifnet *ifp, u_long
NICVF_CORE_LOCK(nic);
if (if_getflags(ifp) & IFF_UP) {
if (if_getdrvflags(ifp) & IFF_DRV_RUNNING) {
- flags = ifp->if_flags ^ nic->if_flags;
- if ((nic->if_flags & ifp->if_flags) &
+ flags = if_getflags(ifp) ^ nic->if_flags;
+ if ((nic->if_flags & if_getflags(ifp)) &
IFF_PROMISC) {
/* Change promiscous mode */
#if 0
@@ -492,7 +492,7 @@ nicvf_if_ioctl(struct ifnet *ifp, u_long
#endif
}
- if ((nic->if_flags ^ ifp->if_flags) &
+ if ((nic->if_flags ^ if_getflags(ifp)) &
IFF_ALLMULTI) {
/* Change multicasting settings */
#if 0
@@ -506,7 +506,7 @@ nicvf_if_ioctl(struct ifnet *ifp, u_long
} else if (if_getdrvflags(ifp) & IFF_DRV_RUNNING)
nicvf_stop_locked(nic);
- nic->if_flags = ifp->if_flags;
+ nic->if_flags = if_getflags(ifp);
NICVF_CORE_UNLOCK(nic);
break;
@@ -528,24 +528,24 @@ nicvf_if_ioctl(struct ifnet *ifp, u_long
break;
case SIOCSIFCAP:
- mask = ifp->if_capenable ^ ifr->ifr_reqcap;
+ mask = if_getcapenable(ifp) ^ ifr->ifr_reqcap;
if (mask & IFCAP_VLAN_MTU) {
/* No work to do except acknowledge the change took. */
- ifp->if_capenable ^= IFCAP_VLAN_MTU;
+ if_togglecapenable(ifp, IFCAP_VLAN_MTU);
}
if (mask & IFCAP_TXCSUM)
- ifp->if_capenable ^= IFCAP_T
Re: svn commit: r313006 - in head: sys/conf sys/libkern sys/libkern/x86 sys/sys tests/sys/kern
On Tue, Jan 31, 2017 at 7:36 AM, Bruce Evans wrote: > On Tue, 31 Jan 2017, Bruce Evans wrote: > Unrolling (or not) may be helpful or harmful for entry and exit code. Helpful, per my earlier benchmarks. > I > think there should by no alignment on entry -- just assume the buffer is > aligned in the usual case, and only run 4% slower when it is misaligned. Please write such a patch and demonstrate the improvement. > The exit code handles up to SHORT * 3 = 768 bytes, not up to 4 or 8 > bytes or up to 3 times that like simpler algorithms. 768 is quite > large, and the exit code is quite slow. It reduces 8 or 4 bytes at a > time without any dependency reduction, and then 1 byte at a time. Yes, this is the important loop to unroll for small inputs. Somehow with the unrolling, it is only ~19% slower than the by-3 algorithm on my system — not 66%. Clang 3.9.1 unrolls both of these trailing loops; here is the first: 0x00401b88 <+584>: cmp$0x38,%rbx 0x00401b8c <+588>: jae0x401b93 0x00401b8e <+590>: mov%rsi,%rdx 0x00401b91 <+593>: jmp0x401be1 0x00401b93 <+595>: lea-0x1(%rdi),%rbx 0x00401b97 <+599>: sub%rdx,%rbx 0x00401b9a <+602>: mov%rsi,%rdx 0x00401b9d <+605>: nopl (%rax) 0x00401ba0 <+608>: crc32q (%rdx),%rax 0x00401ba6 <+614>: crc32q 0x8(%rdx),%rax 0x00401bad <+621>: crc32q 0x10(%rdx),%rax 0x00401bb4 <+628>: crc32q 0x18(%rdx),%rax 0x00401bbb <+635>: crc32q 0x20(%rdx),%rax 0x00401bc2 <+642>: crc32q 0x28(%rdx),%rax 0x00401bc9 <+649>: crc32q 0x30(%rdx),%rax 0x00401bd0 <+656>: crc32q 0x38(%rdx),%rax 0x00401bd7 <+663>: add$0x40,%rdx 0x00401bdb <+667>: add$0x8,%rbx 0x00401bdf <+671>: jne0x401ba0 > I > don't understand the algorithm for joining crcs -- why doesn't it work > to reduce to 12 or 24 bytes in the main loop? It would, but I haven't implemented or tested that. You're welcome to do so and demonstrate an improvement. It does add more lookup table bloat, but perhaps we could just remove the 3x8k table — I'm not sure it adds any benefit over the 3x256 table. > Your benchmarks mainly give results for the <= 768 bytes where most of > the manual optimizations don't apply. 0x000400: asm:68 intrins:62 multitable:684 (ns per buf) 0x000800: asm:132 intrins:133 (ns per buf) 0x002000: asm:449 intrins:446 (ns per buf) 0x008000: asm:1501 intrins:1497 (ns per buf) 0x02: asm:5618 intrins:5609 (ns per buf) (All routines are in a separate compilation unit with no full-program optimization, as they are in the kernel.) > Compiler optimizations are more > likely to help there. So I looked more closely at the last 2 loop. > clang indeed only unrolls the last one, Not in 3.9.1. > only for the unreachable case > with more than 8 bytes on amd64. How is it unreachable? Best, Conrad ___ svn-src-all@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r313020 - head/sys/dev/cxgbe
Author: jhb
Date: Tue Jan 31 18:54:13 2017
New Revision: 313020
URL: https://svnweb.freebsd.org/changeset/base/313020
Log:
Fix a couple of issues with t4iov probe and attach.
- Check for Chelsio vendor ID in probe routines.
- Fail attach instead of faulting if pci_find_dbsf() doesn't find a
device.
PR: 216539
Reported by: asomers
Tested by:Dave Baukus
MFC after:3 days
Sponsored by: Chelsio Communications
Modified:
head/sys/dev/cxgbe/t4_iov.c
Modified: head/sys/dev/cxgbe/t4_iov.c
==
--- head/sys/dev/cxgbe/t4_iov.c Tue Jan 31 16:12:31 2017(r313019)
+++ head/sys/dev/cxgbe/t4_iov.c Tue Jan 31 18:54:13 2017(r313020)
@@ -41,6 +41,7 @@ __FBSDID("$FreeBSD$");
#include
#endif
+#include "common/common.h"
#include "t4_if.h"
struct t4iov_softc {
@@ -106,6 +107,9 @@ t4iov_probe(device_t dev)
uint16_t d;
size_t i;
+ if (pci_get_vendor(dev) != PCI_VENDOR_ID_CHELSIO)
+ return (ENXIO);
+
d = pci_get_device(dev);
for (i = 0; i < nitems(t4iov_pciids); i++) {
if (d == t4iov_pciids[i].device) {
@@ -123,6 +127,9 @@ t5iov_probe(device_t dev)
uint16_t d;
size_t i;
+ if (pci_get_vendor(dev) != PCI_VENDOR_ID_CHELSIO)
+ return (ENXIO);
+
d = pci_get_device(dev);
for (i = 0; i < nitems(t5iov_pciids); i++) {
if (d == t5iov_pciids[i].device) {
@@ -140,6 +147,9 @@ t6iov_probe(device_t dev)
uint16_t d;
size_t i;
+ if (pci_get_vendor(dev) != PCI_VENDOR_ID_CHELSIO)
+ return (ENXIO);
+
d = pci_get_device(dev);
for (i = 0; i < nitems(t6iov_pciids); i++) {
if (d == t6iov_pciids[i].device) {
@@ -161,6 +171,8 @@ t4iov_attach(device_t dev)
sc->sc_main = pci_find_dbsf(pci_get_domain(dev), pci_get_bus(dev),
pci_get_slot(dev), 4);
+ if (sc->sc_main == NULL)
+ return (ENXIO);
if (T4_IS_MAIN_READY(sc->sc_main) == 0)
return (t4iov_attach_child(dev));
return (0);
___
svn-src-all@freebsd.org mailing list
https://lists.freebsd.org/mailman/listinfo/svn-src-all
To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r313021 - head/release
Author: brd (doc,ports committer)
Date: Tue Jan 31 19:09:06 2017
New Revision: 313021
URL: https://svnweb.freebsd.org/changeset/base/313021
Log:
Force a few variables to be upper case when they are evaulated.
This allows specifying, for example: CLOUDWARE=vagrant-vmware
Approved by: re (gjb)
Modified:
head/release/Makefile.vm
Modified: head/release/Makefile.vm
==
--- head/release/Makefile.vmTue Jan 31 18:54:13 2017(r313020)
+++ head/release/Makefile.vmTue Jan 31 19:09:06 2017(r313021)
@@ -58,8 +58,8 @@ cw-${_CW:tl}:
env TARGET=${TARGET} TARGET_ARCH=${TARGET_ARCH} SWAPSIZE=${SWAPSIZE} \
${.CURDIR}/scripts/mk-vmimage.sh \
-C ${.CURDIR}/tools/vmimage.subr -d ${.OBJDIR}/${.TARGET} \
- -i ${.OBJDIR}/${_CW:tl}.img -s ${VMSIZE} -f ${${_CW}_FORMAT} \
- -S ${WORLDDIR} -o ${.OBJDIR}/${${_CW}IMAGE} -c ${${_CW}CONF}
+ -i ${.OBJDIR}/${_CW:tl}.img -s ${VMSIZE} -f ${${_CW:tu}_FORMAT}
\
+ -S ${WORLDDIR} -o ${.OBJDIR}/${${_CW:tu}IMAGE} -c ${${_CW}CONF}
touch ${.TARGET}
cw${_CW:tl}-package:
___
svn-src-all@freebsd.org mailing list
https://lists.freebsd.org/mailman/listinfo/svn-src-all
To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r313022 - svnadmin/conf
Author: glebius Date: Tue Jan 31 19:12:10 2017 New Revision: 313022 URL: https://svnweb.freebsd.org/changeset/base/313022 Log: Add myself for tcpdump import. Modified: svnadmin/conf/sizelimit.conf Modified: svnadmin/conf/sizelimit.conf == --- svnadmin/conf/sizelimit.confTue Jan 31 19:09:06 2017 (r313021) +++ svnadmin/conf/sizelimit.confTue Jan 31 19:12:10 2017 (r313022) @@ -27,3 +27,4 @@ np obrien peter rwatson +glebius ___ svn-src-all@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r313023 - in vendor/tcpdump/4.9.0: . lbl missing tests win32 win32/prj win32/src
Author: glebius Date: Tue Jan 31 19:12:28 2017 New Revision: 313023 URL: https://svnweb.freebsd.org/changeset/base/313023 Log: Import tcpdump-4.9.0. Added: vendor/tcpdump/4.9.0/ vendor/tcpdump/4.9.0/CHANGES vendor/tcpdump/4.9.0/CONTRIBUTING vendor/tcpdump/4.9.0/CREDITS vendor/tcpdump/4.9.0/INSTALL.txt (contents, props changed) vendor/tcpdump/4.9.0/LICENSE vendor/tcpdump/4.9.0/Makefile-devel-adds (contents, props changed) vendor/tcpdump/4.9.0/Makefile.in (contents, props changed) vendor/tcpdump/4.9.0/PLATFORMS vendor/tcpdump/4.9.0/README (contents, props changed) vendor/tcpdump/4.9.0/README.md vendor/tcpdump/4.9.0/Readme.Win32 vendor/tcpdump/4.9.0/VERSION vendor/tcpdump/4.9.0/aclocal.m4 vendor/tcpdump/4.9.0/addrtoname.c (contents, props changed) vendor/tcpdump/4.9.0/addrtoname.h (contents, props changed) vendor/tcpdump/4.9.0/addrtostr.c (contents, props changed) vendor/tcpdump/4.9.0/addrtostr.h (contents, props changed) vendor/tcpdump/4.9.0/af.c (contents, props changed) vendor/tcpdump/4.9.0/af.h (contents, props changed) vendor/tcpdump/4.9.0/ah.h (contents, props changed) vendor/tcpdump/4.9.0/appletalk.h (contents, props changed) vendor/tcpdump/4.9.0/ascii_strcasecmp.c (contents, props changed) vendor/tcpdump/4.9.0/ascii_strcasecmp.h (contents, props changed) vendor/tcpdump/4.9.0/atime.awk (contents, props changed) vendor/tcpdump/4.9.0/atm.h (contents, props changed) vendor/tcpdump/4.9.0/bpf_dump.c (contents, props changed) vendor/tcpdump/4.9.0/chdlc.h (contents, props changed) vendor/tcpdump/4.9.0/checksum.c (contents, props changed) vendor/tcpdump/4.9.0/config.guess vendor/tcpdump/4.9.0/config.h.in (contents, props changed) vendor/tcpdump/4.9.0/config.sub vendor/tcpdump/4.9.0/configure (contents, props changed) vendor/tcpdump/4.9.0/configure.in (contents, props changed) vendor/tcpdump/4.9.0/cpack.c (contents, props changed) vendor/tcpdump/4.9.0/cpack.h (contents, props changed) vendor/tcpdump/4.9.0/ether.h (contents, props changed) vendor/tcpdump/4.9.0/ethertype.h (contents, props changed) vendor/tcpdump/4.9.0/extract.h (contents, props changed) vendor/tcpdump/4.9.0/getopt_long.h (contents, props changed) vendor/tcpdump/4.9.0/gmpls.c (contents, props changed) vendor/tcpdump/4.9.0/gmpls.h (contents, props changed) vendor/tcpdump/4.9.0/gmt2local.c (contents, props changed) vendor/tcpdump/4.9.0/gmt2local.h (contents, props changed) vendor/tcpdump/4.9.0/in_cksum.c (contents, props changed) vendor/tcpdump/4.9.0/install-sh (contents, props changed) vendor/tcpdump/4.9.0/interface.h (contents, props changed) vendor/tcpdump/4.9.0/ip.h (contents, props changed) vendor/tcpdump/4.9.0/ip6.h (contents, props changed) vendor/tcpdump/4.9.0/ipproto.c (contents, props changed) vendor/tcpdump/4.9.0/ipproto.h (contents, props changed) vendor/tcpdump/4.9.0/l2vpn.c (contents, props changed) vendor/tcpdump/4.9.0/l2vpn.h (contents, props changed) vendor/tcpdump/4.9.0/lbl/ vendor/tcpdump/4.9.0/lbl/os-osf4.h (contents, props changed) vendor/tcpdump/4.9.0/lbl/os-solaris2.h (contents, props changed) vendor/tcpdump/4.9.0/lbl/os-sunos4.h (contents, props changed) vendor/tcpdump/4.9.0/lbl/os-ultrix4.h (contents, props changed) vendor/tcpdump/4.9.0/llc.h (contents, props changed) vendor/tcpdump/4.9.0/machdep.c (contents, props changed) vendor/tcpdump/4.9.0/machdep.h (contents, props changed) vendor/tcpdump/4.9.0/makemib (contents, props changed) vendor/tcpdump/4.9.0/mib.h (contents, props changed) vendor/tcpdump/4.9.0/missing/ vendor/tcpdump/4.9.0/missing/datalinks.c (contents, props changed) vendor/tcpdump/4.9.0/missing/dlnames.c (contents, props changed) vendor/tcpdump/4.9.0/missing/getopt_long.c (contents, props changed) vendor/tcpdump/4.9.0/missing/snprintf.c (contents, props changed) vendor/tcpdump/4.9.0/missing/strdup.c (contents, props changed) vendor/tcpdump/4.9.0/missing/strlcat.c (contents, props changed) vendor/tcpdump/4.9.0/missing/strlcpy.c (contents, props changed) vendor/tcpdump/4.9.0/missing/strsep.c (contents, props changed) vendor/tcpdump/4.9.0/mkdep (contents, props changed) vendor/tcpdump/4.9.0/mpls.h (contents, props changed) vendor/tcpdump/4.9.0/nameser.h (contents, props changed) vendor/tcpdump/4.9.0/netdissect-stdinc.h (contents, props changed) vendor/tcpdump/4.9.0/netdissect.c (contents, props changed) vendor/tcpdump/4.9.0/netdissect.h (contents, props changed) vendor/tcpdump/4.9.0/nfs.h (contents, props changed) vendor/tcpdump/4.9.0/nfsfh.h (contents, props changed) vendor/tcpdump/4.9.0/nlpid.c (contents, props changed) vendor/tcpdump/4.9.0/nlpid.h (contents, props changed) vendor/tcpdump/4.9.0/openflow.h (contents, props changed) vendor/tcpdump/4.9.0/ospf.h (contents, props changed) vendor/tcpdump/4.9
svn commit: r313024 - in vendor/tcpdump/dist: . lbl missing tests win32/Include win32/prj win32/src
Author: glebius Date: Tue Jan 31 19:17:06 2017 New Revision: 313024 URL: https://svnweb.freebsd.org/changeset/base/313024 Log: Import tcpdump-4.9.0 into dist. Added: vendor/tcpdump/dist/CONTRIBUTING vendor/tcpdump/dist/PLATFORMS vendor/tcpdump/dist/README (contents, props changed) vendor/tcpdump/dist/addrtostr.c (contents, props changed) vendor/tcpdump/dist/addrtostr.h (contents, props changed) vendor/tcpdump/dist/ascii_strcasecmp.c (contents, props changed) vendor/tcpdump/dist/ascii_strcasecmp.h (contents, props changed) vendor/tcpdump/dist/netdissect-stdinc.h (contents, props changed) vendor/tcpdump/dist/netdissect.c (contents, props changed) vendor/tcpdump/dist/print-hncp.c (contents, props changed) vendor/tcpdump/dist/print-lisp.c (contents, props changed) vendor/tcpdump/dist/print-medsa.c (contents, props changed) vendor/tcpdump/dist/print-nsh.c (contents, props changed) vendor/tcpdump/dist/print-resp.c (contents, props changed) vendor/tcpdump/dist/print-vxlan-gpe.c (contents, props changed) vendor/tcpdump/dist/print.c (contents, props changed) vendor/tcpdump/dist/print.h (contents, props changed) vendor/tcpdump/dist/strtoaddr.c (contents, props changed) vendor/tcpdump/dist/strtoaddr.h (contents, props changed) vendor/tcpdump/dist/tests/OLSRv1_HNA_sgw_1.out vendor/tcpdump/dist/tests/OLSRv1_HNA_sgw_1.pcap (contents, props changed) vendor/tcpdump/dist/tests/RADIUS-port1700.pcap (contents, props changed) vendor/tcpdump/dist/tests/aarp-heapoverflow-1.out vendor/tcpdump/dist/tests/aarp-heapoverflow-1.pcap (contents, props changed) vendor/tcpdump/dist/tests/aarp-heapoverflow-2.out vendor/tcpdump/dist/tests/aarp-heapoverflow-2.pcap (contents, props changed) vendor/tcpdump/dist/tests/arp-too-long-tha.out vendor/tcpdump/dist/tests/arp-too-long-tha.pcap (contents, props changed) vendor/tcpdump/dist/tests/atm-heapoverflow.out vendor/tcpdump/dist/tests/atm-heapoverflow.pcap (contents, props changed) vendor/tcpdump/dist/tests/atm-oam-heapoverflow.out vendor/tcpdump/dist/tests/atm-oam-heapoverflow.pcap (contents, props changed) vendor/tcpdump/dist/tests/bad-ipv4-version-pgm-heapoverflow.out vendor/tcpdump/dist/tests/bad-ipv4-version-pgm-heapoverflow.pcap (contents, props changed) vendor/tcpdump/dist/tests/bfd-raw-auth-md5-v.out vendor/tcpdump/dist/tests/bfd-raw-auth-md5.out vendor/tcpdump/dist/tests/bfd-raw-auth-md5.pcap (contents, props changed) vendor/tcpdump/dist/tests/bfd-raw-auth-sha1-v.out vendor/tcpdump/dist/tests/bfd-raw-auth-sha1.out vendor/tcpdump/dist/tests/bfd-raw-auth-sha1.pcap (contents, props changed) vendor/tcpdump/dist/tests/bfd-raw-auth-simple-v.out vendor/tcpdump/dist/tests/bfd-raw-auth-simple.out vendor/tcpdump/dist/tests/bfd-raw-auth-simple.pcap (contents, props changed) vendor/tcpdump/dist/tests/bgp-aigp.out vendor/tcpdump/dist/tests/bgp-aigp.pcap (contents, props changed) vendor/tcpdump/dist/tests/bgp-large-community.out vendor/tcpdump/dist/tests/bgp-large-community.pcap (contents, props changed) vendor/tcpdump/dist/tests/calm-fast-mac-lookup-heapoverflow.out vendor/tcpdump/dist/tests/calm-fast-mac-lookup-heapoverflow.pcap (contents, props changed) vendor/tcpdump/dist/tests/cve-2014-8767-OLSR.out vendor/tcpdump/dist/tests/cve-2014-8767-OLSR.pcap (contents, props changed) vendor/tcpdump/dist/tests/cve-2014-8768-Geonet.out vendor/tcpdump/dist/tests/cve-2014-8768-Geonet.pcap (contents, props changed) vendor/tcpdump/dist/tests/cve-2014-8769-AODV.out vendor/tcpdump/dist/tests/cve-2014-8769-AODV.pcap (contents, props changed) vendor/tcpdump/dist/tests/dhcp-mud.out vendor/tcpdump/dist/tests/dhcp-mud.pcap (contents, props changed) vendor/tcpdump/dist/tests/dhcpv6-mud.out vendor/tcpdump/dist/tests/dhcpv6-mud.pcap (contents, props changed) vendor/tcpdump/dist/tests/dnssec-vv.out vendor/tcpdump/dist/tests/dnssec.pcap (contents, props changed) vendor/tcpdump/dist/tests/frf15-heapoverflow.out vendor/tcpdump/dist/tests/frf15-heapoverflow.pcap (contents, props changed) vendor/tcpdump/dist/tests/geonet-mac-lookup-heapoverflow.out vendor/tcpdump/dist/tests/geonet-mac-lookup-heapoverflow.pcap (contents, props changed) vendor/tcpdump/dist/tests/gre-heapoverflow-1.out vendor/tcpdump/dist/tests/gre-heapoverflow-1.pcap (contents, props changed) vendor/tcpdump/dist/tests/gre-heapoverflow-2.out vendor/tcpdump/dist/tests/gre-heapoverflow-2.pcap (contents, props changed) vendor/tcpdump/dist/tests/heap-overflow-1.out vendor/tcpdump/dist/tests/heap-overflow-1.pcap (contents, props changed) vendor/tcpdump/dist/tests/heap-overflow-2.out vendor/tcpdump/dist/tests/heap-overflow-2.pcap (contents, props changed) vendor/tcpdump/dist/tests/heapoverflow-EXTRACT_16BITS.out vendor/tcpdump/dist/tests/heapoverflow-EXTRACT_16BITS.pcap (contents, props changed) vendor/tcpdump/dist/tests/he
svn commit: r313025 - head/tests/sys/netinet
Author: asomers
Date: Tue Jan 31 20:13:50 2017
New Revision: 313025
URL: https://svnweb.freebsd.org/changeset/base/313025
Log:
Add tests for multi-fib IPv6 routing
PR: 196361
Submitted by: jhujh...@adjectivism.org
Reported by: Jason Healy
MFC after:4 weeks
Sponsored by: Spectra Logic Corp
Modified:
head/tests/sys/netinet/fibs_test.sh
Modified: head/tests/sys/netinet/fibs_test.sh
==
--- head/tests/sys/netinet/fibs_test.sh Tue Jan 31 19:17:06 2017
(r313024)
+++ head/tests/sys/netinet/fibs_test.sh Tue Jan 31 20:13:50 2017
(r313025)
@@ -75,9 +75,9 @@ arpresolve_checks_interface_fib_body()
get_fibs 2
# Configure TAP interfaces
- setup_tap "$FIB0" ${ADDR0} ${MASK0}
+ setup_tap "$FIB0" inet ${ADDR0} ${MASK0}
TAP0=$TAP
- setup_tap "$FIB1" ${ADDR1} ${MASK1}
+ setup_tap "$FIB1" inet ${ADDR1} ${MASK1}
TAP1=$TAP
# Simulate a crossover cable
@@ -112,7 +112,7 @@ arpresolve_checks_interface_fib_cleanup(
atf_test_case loopback_and_network_routes_on_nondefault_fib cleanup
loopback_and_network_routes_on_nondefault_fib_head()
{
- atf_set "descr" "When creating and deleting loopback routes, use the
interface's fib"
+ atf_set "descr" "When creating and deleting loopback IPv4 routes, use
the interface's fib"
atf_set "require.user" "root"
atf_set "require.config" "fibs"
}
@@ -132,7 +132,7 @@ loopback_and_network_routes_on_nondefaul
get_fibs 1
# Configure a TAP interface
- setup_tap ${FIB0} ${ADDR} ${MASK}
+ setup_tap ${FIB0} inet ${ADDR} ${MASK}
# Check whether the host route exists in only the correct FIB
setfib ${FIB0} netstat -rn -f inet | grep -q "^${ADDR}.*UHS.*lo0"
@@ -156,7 +156,7 @@ loopback_and_network_routes_on_nondefaul
setfib 0 netstat -rn -f inet | \
grep -q "^${SUBNET}/${MASK}.*${TAPD}"
if [ 0 -eq $? ]; then
- setfib ${FIB0} netstat -rn -f inet
+ setfib 0 netstat -rn -f inet
atf_fail "Network route appeared in the wrong FIB"
fi
}
@@ -166,12 +166,70 @@ loopback_and_network_routes_on_nondefaul
cleanup_tap
}
+atf_test_case loopback_and_network_routes_on_nondefault_fib_inet6 cleanup
+loopback_and_network_routes_on_nondefault_fib_inet6_head()
+{
+ atf_set "descr" "When creating and deleting loopback IPv6 routes, use
the interface's fib"
+ atf_set "require.user" "root"
+ atf_set "require.config" "fibs"
+}
+
+loopback_and_network_routes_on_nondefault_fib_inet6_body()
+{
+ atf_expect_fail "PR196361 IPv6 network routes don't respect
net.add_addr_allfibs=0"
+ # Configure the TAP interface to use a nonrouteable RFC3849
+ # address and a non-default fib
+ ADDR="2001:db8::2"
+ SUBNET="2001:db8::"
+ MASK="64"
+
+ # Check system configuration
+ if [ 0 != `sysctl -n net.add_addr_allfibs` ]; then
+ atf_skip "This test requires net.add_addr_allfibs=0"
+ fi
+ get_fibs 1
+
+ # Configure a TAP interface
+ setup_tap ${FIB0} inet6 ${ADDR} ${MASK}
+
+ # Check whether the host route exists in only the correct FIB
+ setfib ${FIB0} netstat -rn -f inet6 | grep -q "^${ADDR}.*UHS.*lo0"
+ if [ 0 -ne $? ]; then
+ setfib ${FIB0} netstat -rn -f inet6
+ atf_fail "Host route did not appear in the correct FIB"
+ fi
+ setfib 0 netstat -rn -f inet6 | grep -q "^${ADDR}.*UHS.*lo0"
+ if [ 0 -eq $? ]; then
+ setfib 0 netstat -rn -f inet6
+ atf_fail "Host route appeared in the wrong FIB"
+ fi
+
+ # Check whether the network route exists in only the correct FIB
+ setfib ${FIB0} netstat -rn -f inet6 | \
+ grep -q "^${SUBNET}/${MASK}.*${TAPD}"
+ if [ 0 -ne $? ]; then
+ setfib ${FIB0} netstat -rn -f inet6
+ atf_fail "Network route did not appear in the correct FIB"
+ fi
+ setfib 0 netstat -rn -f inet6 | \
+ grep -q "^${SUBNET}/${MASK}.*${TAPD}"
+ if [ 0 -eq $? ]; then
+ setfib 0 netstat -rn -f inet6
+ atf_fail "Network route appeared in the wrong FIB"
+ fi
+}
+
+loopback_and_network_routes_on_nondefault_fib_inet6_cleanup()
+{
+ cleanup_tap
+}
+
# Regression test for kern/187552
atf_test_case default_route_with_multiple_fibs_on_same_subnet cleanup
default_route_with_multiple_fibs_on_same_subnet_head()
{
- atf_set "descr" "Multiple interfaces on the same subnet but with
different fibs can both have default routes"
+ atf_set "descr" "Multiple interfaces on the same subnet but with
different fibs can both have default IPv4 routes"
atf_set "require.user" "root"
atf_set "require.config" "fibs"
}
@@ -193,9 +251,9 @@ default_route_with_m
svn commit: r313026 - head/sys/dev/ath
Author: adrian Date: Tue Jan 31 20:57:40 2017 New Revision: 313026 URL: https://svnweb.freebsd.org/changeset/base/313026 Log: [ath] log seqno, type and subtype when assigning sequence numbers for A-MPDU. This is just to improve adrian-debugging. Modified: head/sys/dev/ath/if_ath_tx.c Modified: head/sys/dev/ath/if_ath_tx.c == --- head/sys/dev/ath/if_ath_tx.cTue Jan 31 20:13:50 2017 (r313025) +++ head/sys/dev/ath/if_ath_tx.cTue Jan 31 20:57:40 2017 (r313026) @@ -2979,7 +2979,9 @@ ath_tx_tid_seqno_assign(struct ath_softc M_SEQNO_SET(m0, seqno); /* Return so caller can do something with it if needed */ - DPRINTF(sc, ATH_DEBUG_SW_TX, "%s: -> seqno=%d\n", __func__, seqno); + DPRINTF(sc, ATH_DEBUG_SW_TX, + "%s: -> subtype=0x%x, tid=%d, seqno=%d\n", + __func__, subtype, tid, seqno); return seqno; } ___ svn-src-all@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r313027 - head/release
Author: brd (doc,ports committer)
Date: Tue Jan 31 21:59:10 2017
New Revision: 313027
URL: https://svnweb.freebsd.org/changeset/base/313027
Log:
Add a new vagrant-box target, to make it easier and less redundant to only
produce the Vagrant box and not upload it.
Approved by: re (gjb)
Modified:
head/release/Makefile.vagrant
Modified: head/release/Makefile.vagrant
==
--- head/release/Makefile.vagrant Tue Jan 31 20:57:40 2017
(r313026)
+++ head/release/Makefile.vagrant Tue Jan 31 21:59:10 2017
(r313027)
@@ -24,8 +24,8 @@ VAGRANT_VERSION!= date +%Y.%m.%d
VAGRANT_TARGET:= ${OSRELEASE}${SNAPSHOT_DATE}
.if !empty(CLOUDWARE)
. for _PROVIDER in ${CLOUDWARE}
-. if ${_PROVIDER:MVAGRANT*}
-VAGRANT_PROVIDERS+=${_PROVIDER:S/VAGRANT-//:tl}
+. if ${_PROVIDER:tu:MVAGRANT*}
+VAGRANT_PROVIDERS+=${_PROVIDER:tu:S/VAGRANT-//:tl}
. endif
. endfor
.endif
@@ -70,13 +70,16 @@ ${VAGRANT_TARGET}.${PROVIDER}.box: ${PRO
. endif
CLEANFILES+= vagrant-do-upload-${PROVIDER}
-vagrant-do-upload-${PROVIDER}: ${VAGRANT_TARGET}.${PROVIDER}.box
+vagrant-do-upload-${PROVIDER}: vagrant-box
. if ${PROVIDER} == "virtualbox"
${.CURDIR}/scripts/atlas-upload.sh -b ${TYPE}-${REVISION}-${BRANCH} -f
${VAGRANT_TARGET}.${PROVIDER}.box -p ${PROVIDER} -k ${VAGRANT_KEY} -u
${VAGRANT_USERNAME} -v ${VAGRANT_VERSION}
. elif ${PROVIDER} == "vmware"
${.CURDIR}/scripts/atlas-upload.sh -b ${TYPE}-${REVISION}-${BRANCH} -f
${VAGRANT_TARGET}.${PROVIDER}.box -p ${PROVIDER}_desktop -k ${VAGRANT_KEY} -u
${VAGRANT_USERNAME} -v ${VAGRANT_VERSION}
. endif
touch ${.OBJDIR}/${.TARGET}
+
+vagrant-box: ${VAGRANT_TARGET}.${PROVIDER}.box
+
.endfor
vagrant-upload:${VAGRANT_UPLOAD_TGTS}
___
svn-src-all@freebsd.org mailing list
https://lists.freebsd.org/mailman/listinfo/svn-src-all
To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r313028 - head/sys/net80211
Author: avos
Date: Tue Jan 31 22:21:08 2017
New Revision: 313028
URL: https://svnweb.freebsd.org/changeset/base/313028
Log:
net80211: cleanup ic_flags / iv_flags
Flags
- IEEE80211_F_ASCAN
- IEEE80211_F_SIBSS
- IEEE80211_F_IBSSON
are not used since r170530 (old WI compatibility ioctls removal)
and r178354 (removed from other places).
IEEE80211_F_TXPOW_FIXED was never utilized; initially added with
IEEE80211_F_TXPOW_AUTO / IEEE80211_F_TXPOW_OFF flags,
which were replaced with IEEE80211_C_TXPMGT capability check in r138568.
Reviewed by: adrian
Differential Revision:https://reviews.freebsd.org/D9369
Modified:
head/sys/net80211/ieee80211_node.c
head/sys/net80211/ieee80211_var.h
Modified: head/sys/net80211/ieee80211_node.c
==
--- head/sys/net80211/ieee80211_node.c Tue Jan 31 21:59:10 2017
(r313027)
+++ head/sys/net80211/ieee80211_node.c Tue Jan 31 22:21:08 2017
(r313028)
@@ -349,7 +349,6 @@ ieee80211_create_ibss(struct ieee80211va
ni->ni_fhindex = 1;
}
if (vap->iv_opmode == IEEE80211_M_IBSS) {
- vap->iv_flags |= IEEE80211_F_SIBSS;
ni->ni_capinfo |= IEEE80211_CAPINFO_IBSS; /* XXX */
if (vap->iv_flags & IEEE80211_F_DESBSSID)
IEEE80211_ADDR_COPY(ni->ni_bssid, vap->iv_des_bssid);
Modified: head/sys/net80211/ieee80211_var.h
==
--- head/sys/net80211/ieee80211_var.h Tue Jan 31 21:59:10 2017
(r313027)
+++ head/sys/net80211/ieee80211_var.h Tue Jan 31 22:21:08 2017
(r313028)
@@ -572,8 +572,7 @@ MALLOC_DECLARE(M_80211_VAP);
#defineIEEE80211_F_PRIVACY 0x0010 /* CONF: privacy
enabled */
#defineIEEE80211_F_PUREG 0x0020 /* CONF: 11g w/o 11b
sta's */
#defineIEEE80211_F_SCAN0x0080 /* STATUS: scanning */
-#defineIEEE80211_F_ASCAN 0x0100 /* STATUS: active scan
*/
-#defineIEEE80211_F_SIBSS 0x0200 /* STATUS: start IBSS */
+/* 0x0300 reserved */
/* NB: this is intentionally setup to be IEEE80211_CAPINFO_SHORT_SLOTTIME */
#defineIEEE80211_F_SHSLOT 0x0400 /* STATUS: use short
slot time*/
#defineIEEE80211_F_PMGTON 0x0800 /* CONF: Power mgmt
enable */
@@ -581,8 +580,7 @@ MALLOC_DECLARE(M_80211_VAP);
#defineIEEE80211_F_WME 0x2000 /* CONF: enable WME use
*/
#defineIEEE80211_F_BGSCAN 0x4000 /* CONF: bg scan
enabled (???)*/
#defineIEEE80211_F_SWRETRY 0x8000 /* CONF: sw tx retry
enabled */
-#define IEEE80211_F_TXPOW_FIXED0x0001 /* TX Power: fixed rate
*/
-#defineIEEE80211_F_IBSSON 0x0002 /* CONF: IBSS creation
enable */
+/* 0x0003 reserved */
#defineIEEE80211_F_SHPREAMBLE 0x0004 /* STATUS: use short
preamble */
#defineIEEE80211_F_DATAPAD 0x0008 /* CONF: do alignment
pad */
#defineIEEE80211_F_USEPROT 0x0010 /* STATUS: protection
enabled */
@@ -600,9 +598,9 @@ MALLOC_DECLARE(M_80211_VAP);
#defineIEEE80211_F_DWDS0x8000 /* CONF: Dynamic WDS
enabled */
#defineIEEE80211_F_BITS \
- "\20\1TURBOP\2COMP\3FF\4BURST\5PRIVACY\6PUREG\10SCAN\11ASCAN\12SIBSS" \
- "\13SHSLOT\14PMGTON\15DESBSSID\16WME\17BGSCAN\20SWRETRY\21TXPOW_FIXED" \
- "\22IBSSON\23SHPREAMBLE\24DATAPAD\25USEPROT\26USERBARKER\27CSAPENDING" \
+ "\20\1TURBOP\2COMP\3FF\4BURST\5PRIVACY\6PUREG\10SCAN" \
+ "\13SHSLOT\14PMGTON\15DESBSSID\16WME\17BGSCAN\20SWRETRY" \
+ "\23SHPREAMBLE\24DATAPAD\25USEPROT\26USERBARKER\27CSAPENDING" \
"\30WPA1\31WPA2\32DROPUNENC\33COUNTERM\34HIDESSID\35NOBRIDG\36PCF" \
"\37DOTH\40DWDS"
___
svn-src-all@freebsd.org mailing list
https://lists.freebsd.org/mailman/listinfo/svn-src-all
To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
Re: svn commit: r312988 - in head/sys: compat/cloudabi compat/linux kern sys
On Monday, January 30, 2017 12:57:23 PM Edward Tomasz Napierala wrote: > Author: trasz > Date: Mon Jan 30 12:57:22 2017 > New Revision: 312988 > URL: https://svnweb.freebsd.org/changeset/base/312988 > > Log: > Add kern_listen(), kern_shutdown(), and kern_socket(), and use them > instead of their sys_*() counterparts in various compats. The svr4 > is left untouched, because there's no point. Note that you can compile test svr4 since it is still in the tree. If we want to remove svr4, then we should remove it. However, we should maintain code that is in the tree if it is still there. -- John Baldwin ___ svn-src-all@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
Re: svn commit: r312910 - in head: . etc/etc.pc98 etc/rc.d lib/libsysdecode libexec release release/doc release/doc/en_US.ISO8859-1/hardware release/doc/en_US.ISO8859-1/readme release/doc/share/exampl
On Saturday, January 28, 2017 02:22:15 AM Takahashi Yoshihiro wrote: > Author: nyan > Date: Sat Jan 28 02:22:15 2017 > New Revision: 312910 > URL: https://svnweb.freebsd.org/changeset/base/312910 > > Log: > Remove pc98 support completely. > I thank all developers and contributors for pc98. > > Relnotes: yes BTW, my impression was that there are some other device drivers that are effectively PC-98 only (e.g. everything that uses scsi_low.c) but they might have pccard attachments for use with PC-98 laptops? Perhaps Warner might know? It seems stg(4) had PCI variants, but nsp(4), ncv(4), and stg(4) all came from NetBSD/pc98 via PAO. -- John Baldwin ___ svn-src-all@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
Re: svn commit: r312910 - in head: . etc/etc.pc98 etc/rc.d lib/libsysdecode libexec release release/doc release/doc/en_US.ISO8859-1/hardware release/doc/en_US.ISO8859-1/readme release/doc/share/exampl
On Tue, Jan 31, 2017 at 3:20 PM, John Baldwin wrote: > On Saturday, January 28, 2017 02:22:15 AM Takahashi Yoshihiro wrote: >> Author: nyan >> Date: Sat Jan 28 02:22:15 2017 >> New Revision: 312910 >> URL: https://svnweb.freebsd.org/changeset/base/312910 >> >> Log: >> Remove pc98 support completely. >> I thank all developers and contributors for pc98. >> >> Relnotes: yes > > BTW, my impression was that there are some other device drivers > that are effectively PC-98 only (e.g. everything that uses scsi_low.c) > but they might have pccard attachments for use with PC-98 laptops? > > Perhaps Warner might know? > > It seems stg(4) had PCI variants, but nsp(4), ncv(4), and stg(4) > all came from NetBSD/pc98 via PAO. These all work correctly on any PC Card machine. The only reason they came in this way was because these devices were original marketed only in Japan. I've used all these cards with external SCSI drives in the past. As far as I know, only the if_snc driver, which was removed, is truly pc98 specific. It is wired in such a way that cannot be used in ibm-at compatible laptops. IIRC, it had hard-wired memory decode lines that landed in the middle of the VGA graphics pages or BIOS low memory areas. I have one of these cards still, and it will be detected on my laptops, but can't work due to the required mappings. Now, there's an different question about whether it is time to retire some of the now-ancient SCSI cards from the system, but that's a different kettle of fish that's larger than just nsp, ncv and stg. Warner ___ svn-src-all@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r313029 - svnadmin/conf
Author: matthew (ports committer) Date: Tue Jan 31 22:43:35 2017 New Revision: 313029 URL: https://svnweb.freebsd.org/changeset/base/313029 Log: The Grim Reaper calls. Take in the commit bits for the following people who have neither made a src commit for at least 18 months, nor have they responded to multiple emails about their status as committers. This is clearing up a backlog of about 3 years where the grim reaper process has not been happening. agc Alistair Crooks akiyama Shunsuke Akiyama andre Andre Oppermann charnier Philippe Charnier cherryCherry Mathew das David Schultz davidch David Christensen edavisEric Davis fjoe Max Khon iwasaki Mitsuru IWASAKI kuriyama Jun Kuriyama monthadar Monthadar Jaberi psPaul Saab rnoland Robert Noland sanpeiMIHIRA Yoshiro simon Simon Nielsen stefanf Stefan Farfeleder syuu Takuya ASADA torek Chris Torek venkatVenkat Duvvuru Core would like to thank all of these people for their contributions to the project, and wishes them well in any future endeavours. While here, tidy up the mentors file. In addition to certain of the above who were still technically under mentorship, remove entries for carl, miwi and snb who had previously given up their commit bits. Approved by: core (implicit) Modified: svnadmin/conf/access svnadmin/conf/mentors Modified: svnadmin/conf/access == --- svnadmin/conf/accessTue Jan 31 22:21:08 2017(r313028) +++ svnadmin/conf/accessTue Jan 31 22:43:35 2017(r313029) @@ -21,13 +21,10 @@ ache achim adrian ae -agc -akiyama allanjude alc alfred ambrisko -andre andreast andrew anish @@ -50,16 +47,12 @@ brueffer bryanv bz cem -charnier -cherry cognet cperciva cracauer csjp cy dab -das -davidch davidcs dchagin def @@ -74,14 +67,12 @@ dumbbell dwmalone dwmalone=freebsd-committ...@maths.tcd.ie eadler ed -edavis emaste emax eri erj fabient fanf -fjoe gabor gad gallatin @@ -102,7 +93,6 @@ hselasky ian imp ivadasz -iwasaki jah jamie jasone @@ -135,7 +125,6 @@ ken kevlo kib kp -kuriyama landonf lidl loos @@ -155,7 +144,6 @@ mizhka mjg mm mmel -monthadar mp mr n_hibmadevnull @@ -178,13 +166,11 @@ pho pjd pkelsey pluknet -ps pstef qingli ray remko rmacklem -rnoland roberto rodrigc royger @@ -194,7 +180,6 @@ rrs rstone ru rwatson bb+lists.freebsd.cvs-committ...@cyrus.watson.org -sanpei sbruno scf schweikh @@ -203,23 +188,19 @@ se sephe sgalabov silby -simon sjg skra slm smh sobomax stas -stefanf stevek syrinx -syuu takawata theraven thomas thompsa tijl -torek trasz truckman tsoome @@ -228,7 +209,6 @@ tychon ume uqs vangyzen -venkat whu will wma Modified: svnadmin/conf/mentors == --- svnadmin/conf/mentors Tue Jan 31 22:21:08 2017(r313028) +++ svnadmin/conf/mentors Tue Jan 31 22:43:35 2017(r313029) @@ -12,10 +12,7 @@ # Mentee Mentor Optional comment achim scottl Co-mentor: emaste anish neelCo-mentor: grehan -agcscottl Co-mentor: emax badger kib Co-mentor: vangyzen -carl jimharris -cherry gibbs dabvangyzen defpjd dexuan sephe @@ -27,14 +24,9 @@ jwd rmacklem kadesaiken Co-mentor: scottl, ambrisko karels gnn mahrensmckusick -miwi rwatson -monthadar adrian peterj jhb Co-mentor: grog phil theravenCo-mentor: sjg pstef pfg slmken Co-mentor: scottl, ambrisko -snbdwmalone stevek sjg -torek rpaulo tsoome allanjude Co-mentor: imp -venkat delphij Co-mentor: luigi, jhb ___ svn-src-all@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
Re: svn commit: r312910 - in head: . etc/etc.pc98 etc/rc.d lib/libsysdecode libexec release release/doc release/doc/en_US.ISO8859-1/hardware release/doc/en_US.ISO8859-1/readme release/doc/share/exampl
On Tuesday, January 31, 2017 03:33:55 PM Warner Losh wrote: > On Tue, Jan 31, 2017 at 3:20 PM, John Baldwin wrote: > > On Saturday, January 28, 2017 02:22:15 AM Takahashi Yoshihiro wrote: > >> Author: nyan > >> Date: Sat Jan 28 02:22:15 2017 > >> New Revision: 312910 > >> URL: https://svnweb.freebsd.org/changeset/base/312910 > >> > >> Log: > >> Remove pc98 support completely. > >> I thank all developers and contributors for pc98. > >> > >> Relnotes: yes > > > > BTW, my impression was that there are some other device drivers > > that are effectively PC-98 only (e.g. everything that uses scsi_low.c) > > but they might have pccard attachments for use with PC-98 laptops? > > > > Perhaps Warner might know? > > > > It seems stg(4) had PCI variants, but nsp(4), ncv(4), and stg(4) > > all came from NetBSD/pc98 via PAO. > > These all work correctly on any PC Card machine. The only reason they > came in this way was because these devices were original marketed only > in Japan. I've used all these cards with external SCSI drives in the > past. > > As far as I know, only the if_snc driver, which was removed, is truly > pc98 specific. It is wired in such a way that cannot be used in ibm-at > compatible laptops. IIRC, it had hard-wired memory decode lines that > landed in the middle of the VGA graphics pages or BIOS low memory > areas. I have one of these cards still, and it will be detected on my > laptops, but can't work due to the required mappings. > > Now, there's an different question about whether it is time to retire > some of the now-ancient SCSI cards from the system, but that's a > different kettle of fish that's larger than just nsp, ncv and stg. Fair enough. I haven't fully put away my 12 axe and am toying with dropping any ISA-only storage and NIC drivers (and perhaps pccard-only as well in that case). Hardware that wants to use ISA/pccard for storage is probably happier running 4.x anyway. One question is if we should drop ISA attachments in that case for drivers that support PCI and ISA. However, there's a fair list of ISA-only adapters that would be a good place to start anyway. One concern is to not drop any drivers that are commonly used in emulators or hypervisors (ed(4) comes to mind though I think em(4) is probably available in any modern emulators or hypervisors). -- John Baldwin ___ svn-src-all@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r313030 - head/sys/netinet
Author: tuexen
Date: Tue Jan 31 23:34:02 2017
New Revision: 313030
URL: https://svnweb.freebsd.org/changeset/base/313030
Log:
Remove a duplicate debug statement.
MFC after:1 week
Modified:
head/sys/netinet/sctp_timer.c
Modified: head/sys/netinet/sctp_timer.c
==
--- head/sys/netinet/sctp_timer.c Tue Jan 31 22:43:35 2017
(r313029)
+++ head/sys/netinet/sctp_timer.c Tue Jan 31 23:34:02 2017
(r313030)
@@ -719,13 +719,9 @@ start_again:
if (num_mk) {
SCTPDBG(SCTP_DEBUG_TIMER1, "LAST TSN marked was %x\n",
tsnlast);
- SCTPDBG(SCTP_DEBUG_TIMER1, "Num marked for retransmission was
%d peer-rwd:%ld\n",
- num_mk, (u_long)stcb->asoc.peers_rwnd);
- SCTPDBG(SCTP_DEBUG_TIMER1, "LAST TSN marked was %x\n",
- tsnlast);
- SCTPDBG(SCTP_DEBUG_TIMER1, "Num marked for retransmission was
%d peer-rwd:%d\n",
+ SCTPDBG(SCTP_DEBUG_TIMER1, "Num marked for retransmission was
%d peer-rwd:%u\n",
num_mk,
- (int)stcb->asoc.peers_rwnd);
+ stcb->asoc.peers_rwnd);
}
#endif
*num_marked = num_mk;
___
svn-src-all@freebsd.org mailing list
https://lists.freebsd.org/mailman/listinfo/svn-src-all
To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
Re: svn commit: r312910 - in head: . etc/etc.pc98 etc/rc.d lib/libsysdecode libexec release release/doc release/doc/en_US.ISO8859-1/hardware release/doc/en_US.ISO8859-1/readme release/doc/share/exampl
On Tue, Jan 31, 2017 at 3:46 PM, John Baldwin wrote: > On Tuesday, January 31, 2017 03:33:55 PM Warner Losh wrote: >> On Tue, Jan 31, 2017 at 3:20 PM, John Baldwin wrote: >> > On Saturday, January 28, 2017 02:22:15 AM Takahashi Yoshihiro wrote: >> >> Author: nyan >> >> Date: Sat Jan 28 02:22:15 2017 >> >> New Revision: 312910 >> >> URL: https://svnweb.freebsd.org/changeset/base/312910 >> >> >> >> Log: >> >> Remove pc98 support completely. >> >> I thank all developers and contributors for pc98. >> >> >> >> Relnotes: yes >> > >> > BTW, my impression was that there are some other device drivers >> > that are effectively PC-98 only (e.g. everything that uses scsi_low.c) >> > but they might have pccard attachments for use with PC-98 laptops? >> > >> > Perhaps Warner might know? >> > >> > It seems stg(4) had PCI variants, but nsp(4), ncv(4), and stg(4) >> > all came from NetBSD/pc98 via PAO. >> >> These all work correctly on any PC Card machine. The only reason they >> came in this way was because these devices were original marketed only >> in Japan. I've used all these cards with external SCSI drives in the >> past. >> >> As far as I know, only the if_snc driver, which was removed, is truly >> pc98 specific. It is wired in such a way that cannot be used in ibm-at >> compatible laptops. IIRC, it had hard-wired memory decode lines that >> landed in the middle of the VGA graphics pages or BIOS low memory >> areas. I have one of these cards still, and it will be detected on my >> laptops, but can't work due to the required mappings. >> >> Now, there's an different question about whether it is time to retire >> some of the now-ancient SCSI cards from the system, but that's a >> different kettle of fish that's larger than just nsp, ncv and stg. > > Fair enough. I haven't fully put away my 12 axe and am toying with > dropping any ISA-only storage and NIC drivers (and perhaps pccard-only > as well in that case). Hardware that wants to use ISA/pccard for > storage is probably happier running 4.x anyway. One question is if we > should drop ISA attachments in that case for drivers that support PCI > and ISA. However, there's a fair list of ISA-only adapters that would > be a good place to start anyway. One concern is to not drop any drivers > that are commonly used in emulators or hypervisors (ed(4) comes to mind > though I think em(4) is probably available in any modern emulators or > hypervisors). ISA-only attachments would be a good place to start. My quick glance shows this would be le(4), aha(4), cm(4), mse(4), and joy(4). There's maybe half a dozen EISA-only drivers in the tree that I was planning on retiring in 12, along with EISA. Likewise with the MCA attachments. I know that ed, ep, ex, sn, fe and cs PC Card attachments still work, or did fairly recently. All of those also have ISA attachments. I have no clue if the ISA attachments work, though I still have some ISA systems, I got rid of all my SCSI and Network ISA cards some time ago. I had hoped to get all the pnp autoloading code going so we could remove them from GENERIC before the 11 branch, but keep them as modules. I ran out of time. However, there's still time for 12 though. Warner ___ svn-src-all@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r313031 - head/sys/netinet
Author: tuexen
Date: Tue Jan 31 23:36:31 2017
New Revision: 313031
URL: https://svnweb.freebsd.org/changeset/base/313031
Log:
Take the SCTP common header into account when computing the
space available for chunks. This unbreaks the handling of
ICMPV6 packets indicating "packet too big". It just worked
for IPv4 since we are overbooking for IPv4.
MFC after:1 week
Modified:
head/sys/netinet/sctp_usrreq.c
Modified: head/sys/netinet/sctp_usrreq.c
==
--- head/sys/netinet/sctp_usrreq.c Tue Jan 31 23:34:02 2017
(r313030)
+++ head/sys/netinet/sctp_usrreq.c Tue Jan 31 23:36:31 2017
(r313031)
@@ -108,7 +108,7 @@ sctp_pathmtu_adjustment(struct sctp_tcb
/* Adjust that too */
stcb->asoc.smallest_mtu = nxtsz;
/* now off to subtract IP_DF flag if needed */
- overhead = IP_HDR_SIZE;
+ overhead = IP_HDR_SIZE + sizeof(struct sctphdr);
if (sctp_auth_is_required_chunk(SCTP_DATA,
stcb->asoc.peer_auth_chunks)) {
overhead += sctp_get_auth_chunk_len(stcb->asoc.peer_hmac_id);
}
___
svn-src-all@freebsd.org mailing list
https://lists.freebsd.org/mailman/listinfo/svn-src-all
To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r313032 - head/sys/netinet
Author: tuexen
Date: Wed Feb 1 00:10:29 2017
New Revision: 313032
URL: https://svnweb.freebsd.org/changeset/base/313032
Log:
Ensure that the variable bail is always initialized before used.
MFC after:1 week
Modified:
head/sys/netinet/sctp_output.c
Modified: head/sys/netinet/sctp_output.c
==
--- head/sys/netinet/sctp_output.c Tue Jan 31 23:36:31 2017
(r313031)
+++ head/sys/netinet/sctp_output.c Wed Feb 1 00:10:29 2017
(r313032)
@@ -7730,9 +7730,9 @@ sctp_fill_outqueue(struct sctp_tcb *stcb
/* must make even word boundary */
space_left &= 0xfffc;
strq = stcb->asoc.ss_functions.sctp_ss_select_stream(stcb, net, asoc);
+ giveup = 0;
+ bail = 0;
while ((space_left > 0) && (strq != NULL)) {
- giveup = 0;
- bail = 0;
moved = sctp_move_to_outqueue(stcb, strq, space_left,
frag_point,
&giveup, eeor_mode, &bail, so_locked);
stcb->asoc.ss_functions.sctp_ss_scheduled(stcb, net, asoc,
strq, moved);
___
svn-src-all@freebsd.org mailing list
https://lists.freebsd.org/mailman/listinfo/svn-src-all
To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
Re: svn commit: r313006 - in head: sys/conf sys/libkern sys/libkern/x86 sys/sys tests/sys/kern
On Tue, 31 Jan 2017, Conrad Meyer wrote: On Tue, Jan 31, 2017 at 7:36 AM, Bruce Evans wrote: On Tue, 31 Jan 2017, Bruce Evans wrote: Unrolling (or not) may be helpful or harmful for entry and exit code. Helpful, per my earlier benchmarks. I think there should by no alignment on entry -- just assume the buffer is aligned in the usual case, and only run 4% slower when it is misaligned. Please write such a patch and demonstrate the improvement. I now understand the algorithm. The division into 3 has to keep sequentiality (since CRC input is a bit string), so it doesn't work to separate the memory acces of the 3 crc32's by 8 bytes as in my simple test -- they have to be separated by large amounts. Then recombining requires multiplying the polynomials associated with the CRCs of the higher 2 blocks by X^N and reducing again, where N is large an related to the block size. This is done using a large table for each N needed, and to keep things reasonably simple, only 2 N's are used. The exit code handles up to SHORT * 3 = 768 bytes, not up to 4 or 8 bytes or up to 3 times that like simpler algorithms. 768 is quite large, and the exit code is quite slow. It reduces 8 or 4 bytes at a time without any dependency reduction, and then 1 byte at a time. Yes, this is the important loop to unroll for small inputs. Somehow Not like clang does it. Unrolling is useless without the 3-way blocking. with the unrolling, it is only ~19% slower than the by-3 algorithm on my system ??? not 66%. Clang 3.9.1 unrolls both of these trailing loops; here is the first: Maybe 3.9.1 pessimizes the 3-way loop by unrolling it. This would be fairly easy to do. Just replicate the 3 crc32q's a few times, say 8, and do them in a bad order (3 blocks of 8 dependent ones instead of 8 blocks of 3 independent ones). With enough replication, the code would be too large for the hardware to reorder. Inline asm has another advantage here -- volatile on it prevents reordering and might prevent unrolling. Maybe 3.9.1 unpessimizes the inline asms. But I suspect not getting the 3 times speedup is for another reason. 0x00401b88 <+584>: cmp$0x38,%rbx 0x00401b8c <+588>: jae0x401b93 0x00401b8e <+590>: mov%rsi,%rdx 0x00401b91 <+593>: jmp0x401be1 0x00401b93 <+595>: lea-0x1(%rdi),%rbx 0x00401b97 <+599>: sub%rdx,%rbx 0x00401b9a <+602>: mov%rsi,%rdx 0x00401b9d <+605>: nopl (%rax) 0x00401ba0 <+608>: crc32q (%rdx),%rax 0x00401ba6 <+614>: crc32q 0x8(%rdx),%rax 0x00401bad <+621>: crc32q 0x10(%rdx),%rax 0x00401bb4 <+628>: crc32q 0x18(%rdx),%rax 0x00401bbb <+635>: crc32q 0x20(%rdx),%rax 0x00401bc2 <+642>: crc32q 0x28(%rdx),%rax 0x00401bc9 <+649>: crc32q 0x30(%rdx),%rax 0x00401bd0 <+656>: crc32q 0x38(%rdx),%rax 0x00401bd7 <+663>: add$0x40,%rdx 0x00401bdb <+667>: add$0x8,%rbx 0x00401bdf <+671>: jne0x401ba0 No, this unrolling is useless. The crc32q's are dependent on each other, so they take 3 cycles each. There are spare resources to run about 12 instructions during that time. Loop control only takes 3. I don't understand the algorithm for joining crcs -- why doesn't it work to reduce to 12 or 24 bytes in the main loop? It would, but I haven't implemented or tested that. You're welcome to do so and demonstrate an improvement. It does add more lookup table bloat, but perhaps we could just remove the 3x8k table ??? I'm not sure it adds any benefit over the 3x256 table. Your benchmarks mainly give results for the <= 768 bytes where most of the manual optimizations don't apply. Actually, they test only the large buffer case. They used buffer size of 1M and 1k and didn't do the entry and exit code that usually dominates for small buffers. I re-tested with the correct blocking. This was about 10% slower (0.34 -> 0.37 seconds for 10GB), except for clang without intrinsics it was 20% slower (0.43 -> 0.51) seconds. 0x000400: asm:68 intrins:62 multitable:684 (ns per buf) I don't see any signs of this in my test: - a single crc32q in a (C) loop doesn't benefit from unrolling or lose to the extra clang instructions without intrinsics. clang-3.9.0 unrolls this 8-way in the simpler environment of my test program, but this makes no difference. - similarly for a single crc32b in a loop, except when I forgot to change the type of the crc accumulator from uint64_t to uint32_t, gcc was 1 cycle slower in the loop (3 instead of 4). gcc generates an extra instruction to zero-extend the crc, and this is more expensive than usual since it gives gives another dependency. clang optimizes this away. 0x000800: asm:132 intrins:133 (ns per buf) 0x002000: asm:449 intrins:446 (ns per buf) 0x008000: asm:1501 intrins:1497 (ns per buf) 0x02: asm:5618 int
svn commit: r313033 - head/share/misc
Author: jmd (ports committer) Date: Wed Feb 1 01:25:30 2017 New Revision: 313033 URL: https://svnweb.freebsd.org/changeset/base/313033 Log: Add myself (jmd) to committers-ports.dot. Document rene and swills as my mentors. Reviewed by: rene (mentor) Approved by: rene (mentor) Differential Revision:https://reviews.freebsd.org/D9393 Modified: head/share/misc/committers-ports.dot Modified: head/share/misc/committers-ports.dot == --- head/share/misc/committers-ports.dotWed Feb 1 00:10:29 2017 (r313032) +++ head/share/misc/committers-ports.dotWed Feb 1 01:25:30 2017 (r313033) @@ -125,6 +125,7 @@ jgh [label="Jason Helfman\njgh@FreeBSD.o jhale [label="Jason E. Hale\njh...@freebsd.org\n2012/09/10"] jkim [label="Jung-uk Kim\nj...@freebsd.org\n2007/09/12"] jlaffaye [label="Julien Laffaye\njlaff...@freebsd.org\n2011/06/06"] +jmd [label="Johannes M. Dieterich\n...@freebsd.org\n2017/01/09"] jmelo [label="Jean Milanez Melo\njm...@freebsd.org\n2006/03/31"] joerg [label="Joerg Wunsch\njo...@freebsd.org\n1994/08/22"] johans [label="Johan Selst\njoh...@freebsd.org\n2006/04/01"] @@ -562,6 +563,7 @@ rene -> bar rene -> cmt rene -> crees rene -> jgh +rene -> jmd rene -> ler rene -> olivierd @@ -594,6 +596,7 @@ stas -> araujo steve -> netchild swills -> feld +swills -> jmd swills -> jrm swills -> milki swills -> pclin ___ svn-src-all@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r313034 - head/sys/dev/usb
Author: imp Date: Wed Feb 1 02:00:18 2017 New Revision: 313034 URL: https://svnweb.freebsd.org/changeset/base/313034 Log: Add USB IDs supported by TI / Intel USB 3410 family (modems of various flavors). This includes both the release products (from Honeywell, Moxa and Multitech) as well as development kits. Modified: head/sys/dev/usb/usbdevs Modified: head/sys/dev/usb/usbdevs == --- head/sys/dev/usb/usbdevsWed Feb 1 01:25:30 2017(r313033) +++ head/sys/dev/usb/usbdevsWed Feb 1 02:00:18 2017(r313034) @@ -587,6 +587,7 @@ vendor CCYU 0x1065 CCYU Technology vendor CURITEL 0x106c Curitel Communications Inc vendor SILABS2 0x10a6 SILABS2 vendor USI 0x10ab USI +vendor HONEYWELL 0x10ac Honeywell vendor LIEBERT20x10af Liebert vendor PLX 0x10b5 PLX vendor ASANTE 0x10bd Asante @@ -594,6 +595,7 @@ vendor SILABS 0x10c4 Silicon Labs vendor SILABS3 0x10c5 Silicon Labs vendor SILABS4 0x10ce Silicon Labs vendor ACTIONS 0x10d6 Actions +vendor MOXA0x110a Moxa vendor ANALOG 0x1110 Analog Devices vendor TENX0x1130 Ten X Technology, Inc. vendor ISSC0x1131 Integrated System Solution Corp. @@ -698,6 +700,7 @@ vendor ELV 0x18ef ELV vendor LINKSYS30x1915 Linksys vendor QUALCOMMINC 0x19d2 Qualcomm, Incorporated vendor QUALCOMM3 0x19f5 Qualcomm, Inc. +vendor ABBOTT 0x1a61 Abbott Diabetics vendor BAYER 0x1a79 Bayer vendor WCH20x1a86 QinHeng Electronics vendor STELERA 0x1a8d Stelera Wireless @@ -806,6 +809,10 @@ product 3COMUSR USRISDN0x008f 3Com U.S product 3COMUSR HOMECONN 0x009d 3Com HomeConnect Camera product 3COMUSR USR56K 0x3021 U.S. Robotics 56000 Voice FaxModem Pro +/* Abbott Diabetics */ +product ABBOTT STEREO_PLUG 0x3410 Abbott Diabetics Stereo Plug +product ABBOTT STRIP_PORT 0x3420 Abbott Diabetics Strip Port + /* AboCom products */ product ABOCOM XX1 0x110c XX1 product ABOCOM XX2 0x200c XX2 @@ -2290,6 +2297,9 @@ product HITACHI DVDCAM_USB0x001e DVDCAM /* Holtek products */ product HOLTEK F85 0xa030 Holtek USB gaming keyboard +/* Honeywell */ +product HONEYWELL HGI800x0102 Honeywell HGI80 Wireless USB Gateway + /* HP products */ product HP 895C0x0004 DeskJet 895C product HP 4100C 0x0101 Scanjet 4100C @@ -2448,6 +2458,9 @@ product HUAWEI3COM WUB320G0x0009 Aolynk /* IBM Corporation */ product IBM USBCDROMDRIVE 0x4427 USB CD-ROM Drive +product IBM USB45430x4543 TI IBM USB 4543 Modem +product IBM USB454B0x454b TI IBM USB 454B Modem +product IBM USB454C0x454c TI IBM USB 454C Modem /* Icom products */ product ICOM SP1 0x0004 FTDI compatible adapter @@ -3222,12 +3235,27 @@ product MOTOROLA2 MB886 0x710f MB886 Mo product MOTOROLA4 RT2770 0x9031 RT2770 product MOTOROLA4 RT3070 0x9032 RT3070 +/* Moxa */ +product MOXA MXU1_1110 0x1110 Moxa Uport 1110 +product MOXA MXU1_1130 0x1130 Moxa Uport 1130 +product MOXA MXU1_1131 0x1131 Moxa Uport 1131 +product MOXA MXU1_1150 0x1150 Moxa Uport 1150 +product MOXA MXU1_1151 0x1151 Moxa Uport 1151 + /* MpMan products */ product MPMAN MPF400_2 0x25a8 MPF400 Music Player 2Go product MPMAN MPF400_1 0x36d0 MPF400 Music Player 1Go /* MultiTech products */ +product MULTITECH MT9234ZBA_2 0x0319 MT9234ZBA USB modem (alt) product MULTITECH ATLAS0xf101 MT5634ZBA-USB modem +product MULTITECH GSM 0xf108 GSM USB Modem +product MULTITECH CDMA 0xf109 CDMA USB Modem +product MULTITECH CDMA_FW 0xf110 CDMA USB Modem firmware running +productMULTITECH GSM_FW0xf111 GSM USB Modem firmware running +product MULTITECH EDGE 0xf112 Edge USB Modem +product MULTITECH MT9234MU 0xf114 MT9234 MU +product MULTITECH MT9234ZBA0xf115 MT9234 ZBA /* Mustek products */ product MUSTEK 1200CU 0x0001 1200 CU scanner @@ -4466,6 +4494,13 @@ product TENX UAUDIO0 0xf211 USB audio h /* Texas Intel products */ product TI UTUSB41 0x1446 UT-USB41 hub product TI TUSB20460x2046 TUSB2046 hub +product TI USB3410 0x3410 TI USB 3410 Modem +product TI USB5052 0x5052 TI USB 5250 Modem +product TI FRI20x5053 TI Fish River Island II +product TI USB5052_EEPROM 0x505a TI USB 5250 Modem w/Eeprom +product TI USB5052_FW 0x505f TI USB 5250 Modme w/Firmware running +product TI USB5152 0x5152 TI USB 5152 Modem +product TI EZ430
svn commit: r313035 - head/contrib/binutils/opcodes
Author: jhibbits
Date: Wed Feb 1 02:42:45 2017
New Revision: 313035
URL: https://svnweb.freebsd.org/changeset/base/313035
Log:
Add rfdi opcode to binutils
The rfdi instruction is part of the e500mc and derivative cores. It came into
binutils in a GPLv3 patch, along with the rest of the e500mc instruction set.
Currently only rfdi is planned to be used, so rather than attempt to backport
the full patch, take a surgical route and add instructions as needed.
MFC after:2 weeks
Modified:
head/contrib/binutils/opcodes/ppc-opc.c
Modified: head/contrib/binutils/opcodes/ppc-opc.c
==
--- head/contrib/binutils/opcodes/ppc-opc.c Wed Feb 1 02:00:18 2017
(r313034)
+++ head/contrib/binutils/opcodes/ppc-opc.c Wed Feb 1 02:42:45 2017
(r313035)
@@ -2789,6 +2789,7 @@ const struct powerpc_opcode powerpc_opco
{ "crnor", XL(19,33),XL_MASK,COM,{ BT, BA, BB }
},
{ "rfmci",X(19,38), 0x,PPCRFMCI, { 0 } },
+{ "rfdi",XL(19,39),0x, BOOKE, { 0 } },
{ "rfi", XL(19,50),0x, COM,{ 0 } },
{ "rfci",XL(19,51),0x, PPC403 | BOOKE, { 0 } },
___
svn-src-all@freebsd.org mailing list
https://lists.freebsd.org/mailman/listinfo/svn-src-all
To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
Re: svn commit: r313006 - in head: sys/conf sys/libkern sys/libkern/x86 sys/sys tests/sys/kern
Another reply to this... On Tue, 31 Jan 2017, Conrad Meyer wrote: On Tue, Jan 31, 2017 at 7:36 AM, Bruce Evans wrote: On Tue, 31 Jan 2017, Bruce Evans wrote: Unrolling (or not) may be helpful or harmful for entry and exit code. Helpful, per my earlier benchmarks. I think there should by no alignment on entry -- just assume the buffer is aligned in the usual case, and only run 4% slower when it is misaligned. Please write such a patch and demonstrate the improvement. It is easy to demonstrate. I just put #if 0 around the early alignment code. The result seem too good to be true, so maybe I missed some later dependency on alignment of the addresses: - for 128-byte buffers and misalignment of 3, 10g takes 1.48 seconds with alignment and 1.02 seconds without alignment. This actually makes sense, 128 bytes can be done with 16 8-byte unaligned crc32q's. The alignment code makes it do 15 * 8-but and (5 + 3) * 1-byte. 7 more 3-cycle instructions and overhead too is far more than the cost of letting the CPU do read-combining. - for 4096-byte buffers, the difference is insignificant (0.47 seconds for 10g. I don't understand the algorithm for joining crcs -- why doesn't it work to reduce to 12 or 24 bytes in the main loop? It would, but I haven't implemented or tested that. You're welcome to do so and demonstrate an improvement. It does add more lookup table bloat, but perhaps we could just remove the 3x8k table ??? I'm not sure it adds any benefit over the 3x256 table. Good idea, but the big table is useful. Ifdefing out the LONG case reduces the speed for large buffers from ~0.35 seconds to ~0.43 seconds in the setup below. Ifdefing out the SHORT case only reduces to ~0.39 seconds. I hoped that an even shorter SHORT case would work. I think it now handles 768 bytes (3 * SHORT) in the inner loop. That is 32 sets of 3 crc32q's, and I would have thought that update at the end would take about as long as 1 iteration (3%), but it apparently takes 33%. ... Your benchmarks mainly give results for the <= 768 bytes where most of the manual optimizations don't apply. 0x000400: asm:68 intrins:62 multitable:684 (ns per buf) 0x000800: asm:132 intrins:133 (ns per buf) 0x002000: asm:449 intrins:446 (ns per buf) 0x008000: asm:1501 intrins:1497 (ns per buf) 0x02: asm:5618 intrins:5609 (ns per buf) (All routines are in a separate compilation unit with no full-program optimization, as they are in the kernel.) These seem slow. I modified my program to test the actual kernel code, and get for 10gB on freefall's Xeon (main times in seconds): 0x08: asm(rm):3.41 asm(r):3.07 intrins:6.01 gcc:3.74 (3S = 2.4ns/buf) 0x10: asm(rm):2.05 asm(r):1.70 intrins:2.92 gcc:2.62 (2S = 3/2ns/buf) 0x20: asm(rm):1.63 asm(r):1.58 intrins:1.62 gcc:1.61 (1.6S = 5.12ns/buf) 0x40: asm(rm):1.07 asm(r):1.11 intrins:1.06 gcc:1.14 (1.1S = 7.04ns/buf) 0x80: asm(rm):1.02 asm(r):1.04 intrins:1.03 gcc:1.04 (1.02S = 13.06ns/buf) 0x000100: asm(rm):1.02 asm(r):1.02 intrins:1.02 gcc:1.08 (1.02S = 52.22ns/buf) 0x000200: asm(rm):1.02 asm(r):1.02 intrins:1.02 gcc:1.02 (1.02S = 104.45ns/buf) 0x000400: asm(rm):0.58 asm(r):0.57 intrins:0.57 gcc:0.57 (.57S = 116.43ns/buf) 0x001000: asm(rm):0.62 asm(r):0.57 intrins:0.57 gcc:0.57 (.57S = 233.44ns/buf) 0x002000: asm(rm):0.48 asm(r):0.46 intrins:0.46 gcc:0.46 (.46S = 376.83ns/buf) 0x004000: asm(rm):0.49 asm(r):0.46 intrins:0.46 gcc:0.46 (.46S = 753.66ns/buf) 0x008000: asm(rm):0.49 asm(r):0.38 intrins:0.38 gcc:0.38 (.38S = 1245.18ns/buf) 0x01: asm(rm):0.47 asm(r):0.38 intrins:0.36 gcc:0.38 (.36S = 2359.30ns/buf) 0x02: asm(rm):0.43 asm(r):1.05 intrins:0.35 gcc:0.36 (.35S = 4587.52ns/buf) asm(r) is a fix for clang's slownes with inline asms. Just change the constraint from "rm" to "r". This takes an extra register, but no more uops. This is for the aligned case with no hacks. intrins does something bad for small buffers. Probably just the branch over the dead unrolling. Twice 2.4ns/buf for 8-byte buffers is still very fast. This is 16 cycles. 3 cycles to do 1 crc32q and the rest mainly for 1 function call and too many branches. Bruce___ svn-src-all@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r313036 - in head/sys/powerpc: booke include
Author: jhibbits Date: Wed Feb 1 03:29:13 2017 New Revision: 313036 URL: https://svnweb.freebsd.org/changeset/base/313036 Log: Add Book-E Enhanced Debug (E.D) profile debug support Freescale added the E.D profile to e500mc and derivative cores. From Freescale's EREF reference manual this is enabled by a bit in HID0 and should otherwise default to traditional debug. However, none of the Freescale cores support that bit, and instead always use E.D. This results in kernel panics using the standard debug on e500mc+ cores. Enhanced debug allows debugging of interrupts, including critical interrupts, as it uses a different save/restore registers (srr*). At this time we don't use this ability, so instead share the core of the debug handler code between both handlers. MFC after:3 weeks Modified: head/sys/powerpc/booke/booke_machdep.c head/sys/powerpc/booke/trap_subr.S head/sys/powerpc/include/spr.h Modified: head/sys/powerpc/booke/booke_machdep.c == --- head/sys/powerpc/booke/booke_machdep.c Wed Feb 1 02:42:45 2017 (r313035) +++ head/sys/powerpc/booke/booke_machdep.c Wed Feb 1 03:29:13 2017 (r313036) @@ -187,6 +187,7 @@ extern void *int_watchdog; extern void *int_data_tlb_error; extern void *int_inst_tlb_error; extern void *int_debug; +extern void *int_debug_ed; extern void *int_vec; extern void *int_vecast; #ifdef HWPMC_HOOKS @@ -242,6 +243,7 @@ ivor_setup(void) case FSL_E500mc: case FSL_E5500: SET_TRAP(SPR_IVOR7, int_fpu); + SET_TRAP(SPR_IVOR15, int_debug_ed); break; case FSL_E500v1: case FSL_E500v2: Modified: head/sys/powerpc/booke/trap_subr.S == --- head/sys/powerpc/booke/trap_subr.S Wed Feb 1 02:42:45 2017 (r313035) +++ head/sys/powerpc/booke/trap_subr.S Wed Feb 1 03:29:13 2017 (r313036) @@ -794,6 +794,22 @@ interrupt_vector_top: INTERRUPT(int_debug) STANDARD_CRIT_PROLOG(SPR_SPRG2, PC_BOOKE_CRITSAVE, SPR_CSRR0, SPR_CSRR1) FRAME_SETUP(SPR_SPRG2, PC_BOOKE_CRITSAVE, EXC_DEBUG) + bl int_debug_int + FRAME_LEAVE(SPR_CSRR0, SPR_CSRR1) + rfci + +INTERRUPT(int_debug_ed) + STANDARD_CRIT_PROLOG(SPR_SPRG2, PC_BOOKE_CRITSAVE, SPR_DSRR0, SPR_DSRR1) + FRAME_SETUP(SPR_SPRG2, PC_BOOKE_CRITSAVE, EXC_DEBUG) + bl int_debug_int + FRAME_LEAVE(SPR_DSRR0, SPR_DSRR1) + rfdi + /* .long 0x4c4e */ + +/* Internal helper for debug interrupt handling. */ +/* Common code between e500v1/v2 and e500mc-based cores. */ +int_debug_int: + mflr%r14 GET_CPUINFO(%r3) lwz %r3, (PC_BOOKE_CRITSAVE+CPUSAVE_SRR0)(%r3) bl 0f @@ -819,7 +835,8 @@ INTERRUPT(int_debug) mtspr SPR_SRR0, %r3 lwz %r4, (PC_BOOKE_CRITSAVE+CPUSAVE_SRR1+8)(%r4); mtspr SPR_SRR1, %r4 - b 9f + mtlr%r14 + blr 1: addi%r3, %r1, 8 bl CNAME(trap) @@ -828,10 +845,6 @@ INTERRUPT(int_debug) * We actually need to return to the process with an rfi. */ b trapexit -9: - FRAME_LEAVE(SPR_CSRR0, SPR_CSRR1) - rfci - /* * Common trap code Modified: head/sys/powerpc/include/spr.h == --- head/sys/powerpc/include/spr.h Wed Feb 1 02:42:45 2017 (r313035) +++ head/sys/powerpc/include/spr.h Wed Feb 1 03:29:13 2017 (r313036) @@ -671,6 +671,8 @@ #defineSPR_CSRR1 0x03b /* ..8 59 Critical SRR1 */ #defineSPR_MCSRR0 0x23a /* ..8 570 Machine check SRR0 */ #defineSPR_MCSRR1 0x23b /* ..8 571 Machine check SRR1 */ +#defineSPR_DSRR0 0x23e /* ..8 574 Debug SRR0 */ +#defineSPR_DSRR1 0x23f /* ..8 575 Debug SRR1 */ #defineSPR_MMUCR 0x3b2 /* 4.. MMU Control Register */ #define MMUCR_SWOA(0x8000 >> 7) ___ svn-src-all@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r313037 - in head/sys: amd64/include kern mips/include net powerpc/include sparc64/include
Author: jah
Date: Wed Feb 1 03:32:49 2017
New Revision: 313037
URL: https://svnweb.freebsd.org/changeset/base/313037
Log:
Implement get_pcpu() for the remaining architectures and use it to
replace pcpu_find(curcpu) in MI code.
Modified:
head/sys/amd64/include/pcpu.h
head/sys/kern/kern_rmlock.c
head/sys/mips/include/pcpu.h
head/sys/net/netisr.c
head/sys/powerpc/include/cpufunc.h
head/sys/powerpc/include/pcpu.h
head/sys/sparc64/include/pcpu.h
Modified: head/sys/amd64/include/pcpu.h
==
--- head/sys/amd64/include/pcpu.h Wed Feb 1 03:29:13 2017
(r313036)
+++ head/sys/amd64/include/pcpu.h Wed Feb 1 03:32:49 2017
(r313037)
@@ -78,6 +78,7 @@
extern struct pcpu *pcpup;
+#defineget_pcpu() (pcpup)
#definePCPU_GET(member)(pcpup->pc_ ## member)
#definePCPU_ADD(member, val) (pcpup->pc_ ## member += (val))
#definePCPU_INC(member)PCPU_ADD(member, 1)
@@ -203,6 +204,15 @@ extern struct pcpu *pcpup;
} \
}
+#defineget_pcpu() __extension__ ({
\
+ struct pcpu *__pc; \
+ \
+ __asm __volatile("movq %%gs:%1,%0" \
+ : "=r" (__pc) \
+ : "m" (*(struct pcpu *)(__pcpu_offset(pc_prvspace; \
+ __pc; \
+})
+
#definePCPU_GET(member)__PCPU_GET(pc_ ## member)
#definePCPU_ADD(member, val) __PCPU_ADD(pc_ ## member, val)
#definePCPU_INC(member)__PCPU_INC(pc_ ## member)
Modified: head/sys/kern/kern_rmlock.c
==
--- head/sys/kern/kern_rmlock.c Wed Feb 1 03:29:13 2017(r313036)
+++ head/sys/kern/kern_rmlock.c Wed Feb 1 03:32:49 2017(r313037)
@@ -156,7 +156,7 @@ unlock_rm(struct lock_object *lock)
*/
critical_enter();
td = curthread;
- pc = pcpu_find(curcpu);
+ pc = get_pcpu();
for (queue = pc->pc_rm_queue.rmq_next;
queue != &pc->pc_rm_queue; queue = queue->rmq_next) {
tracker = (struct rm_priotracker *)queue;
@@ -258,7 +258,7 @@ rm_cleanIPI(void *arg)
struct rmlock *rm = arg;
struct rm_priotracker *tracker;
struct rm_queue *queue;
- pc = pcpu_find(curcpu);
+ pc = get_pcpu();
for (queue = pc->pc_rm_queue.rmq_next; queue != &pc->pc_rm_queue;
queue = queue->rmq_next) {
@@ -355,7 +355,7 @@ _rm_rlock_hard(struct rmlock *rm, struct
struct pcpu *pc;
critical_enter();
- pc = pcpu_find(curcpu);
+ pc = get_pcpu();
/* Check if we just need to do a proper critical_exit. */
if (!CPU_ISSET(pc->pc_cpuid, &rm->rm_writecpus)) {
@@ -416,7 +416,7 @@ _rm_rlock_hard(struct rmlock *rm, struct
}
critical_enter();
- pc = pcpu_find(curcpu);
+ pc = get_pcpu();
CPU_CLR(pc->pc_cpuid, &rm->rm_writecpus);
rm_tracker_add(pc, tracker);
sched_pin();
@@ -641,7 +641,7 @@ _rm_rlock_debug(struct rmlock *rm, struc
#ifdef INVARIANTS
if (!(rm->lock_object.lo_flags & LO_RECURSABLE) && !trylock) {
critical_enter();
- KASSERT(rm_trackers_present(pcpu_find(curcpu), rm,
+ KASSERT(rm_trackers_present(get_pcpu(), rm,
curthread) == 0,
("rm_rlock: recursed on non-recursive rmlock %s @ %s:%d\n",
rm->lock_object.lo_name, file, line));
@@ -771,7 +771,7 @@ _rm_assert(const struct rmlock *rm, int
}
critical_enter();
- count = rm_trackers_present(pcpu_find(curcpu), rm, curthread);
+ count = rm_trackers_present(get_pcpu(), rm, curthread);
critical_exit();
if (count == 0)
@@ -797,7 +797,7 @@ _rm_assert(const struct rmlock *rm, int
rm->lock_object.lo_name, file, line);
critical_enter();
- count = rm_trackers_present(pcpu_find(curcpu), rm, curthread);
+ count = rm_trackers_present(get_pcpu(), rm, curthread);
critical_exit();
if (count != 0)
Modified: head/sys/mips/include/pcpu.h
==
--- head/sys/mips/include/pcpu.hWed Feb 1 03:29:13 2017
(r313036)
+++ head/sys/mips/include/pcpu.hWed Feb 1 03:32:49 2017
(r313037)
@@ -65,6 +65,7 @@ extern char pcpu_space[MAXCP
Re: svn commit: r313006 - in head: sys/conf sys/libkern sys/libkern/x86 sys/sys tests/sys/kern
On Tue, Jan 31, 2017 at 7:16 PM, Bruce Evans wrote: > Another reply to this... > > On Tue, 31 Jan 2017, Conrad Meyer wrote: > >> On Tue, Jan 31, 2017 at 7:36 AM, Bruce Evans wrote: >>> >>> On Tue, 31 Jan 2017, Bruce Evans wrote: >>> I >>> think there should by no alignment on entry -- just assume the buffer is >>> aligned in the usual case, and only run 4% slower when it is misaligned. >> >> >> Please write such a patch and demonstrate the improvement. > > > It is easy to demonstrate. I just put #if 0 around the early alignment > code. The result seem too good to be true, so maybe I missed some > later dependency on alignment of the addresses: > - for 128-byte buffers and misalignment of 3, 10g takes 1.48 seconds with > alignment and 1.02 seconds without alignment. > This actually makes sense, 128 bytes can be done with 16 8-byte unaligned > crc32q's. The alignment code makes it do 15 * 8-but and (5 + 3) * 1-byte. > 7 more 3-cycle instructions and overhead too is far more than the cost > of letting the CPU do read-combining. > - for 4096-byte buffers, the difference is insignificant (0.47 seconds for > 10g. I believe it, especially for newer amd64. I seem to recall that older x86 machines had a higher misalignment penalty, but it was largely reduced in (?)Nehalem. Why don't you go ahead and commit that change? >> perhaps we could just remove the 3x8k table — I'm not sure >> it adds any benefit over the 3x256 table. > > > Good idea, but the big table is useful. Ifdefing out the LONG case reduces > the speed for large buffers from ~0.35 seconds to ~0.43 seconds in the > setup below. Ifdefing out the SHORT case only reduces to ~0.39 seconds. Interesting. > I hoped that an even shorter SHORT case would work. I think it now handles > 768 bytes (3 * SHORT) in the inner loop. Right. > That is 32 sets of 3 crc32q's, > and I would have thought that update at the end would take about as long > as 1 iteration (3%), but it apparently takes 33%. The update at the end may be faster with PCLMULQDQ. There are versions of this algorithm that use that in place of the lookup-table combine (for example, Linux has a permissively licensed implementation here: http://lxr.free-electrons.com/source/arch/x86/crypto/crc32c-pcl-intel-asm_64.S ). Unfortunately, PCLMULQDQ uses FPU state, which is inappropriate most of the time in kernel mode. It could be used opportunistically if the thread is already in FPU-save mode or if the input is "big enough" to make it worth it. >>> Your benchmarks mainly give results for the <= 768 bytes where most of >>> the manual optimizations don't apply. >> >> >> 0x000400: asm:68 intrins:62 multitable:684 (ns per buf) >> 0x000800: asm:132 intrins:133 (ns per buf) >> 0x002000: asm:449 intrins:446 (ns per buf) >> 0x008000: asm:1501 intrins:1497 (ns per buf) >> 0x02: asm:5618 intrins:5609 (ns per buf) >> >> (All routines are in a separate compilation unit with no full-program >> optimization, as they are in the kernel.) > > > These seem slow. I modified my program to test the actual kernel code, > and get for 10gB on freefall's Xeon (main times in seconds): Freefall has a Haswell Xeon @ 3.3GHz. My laptop is a Sandy Bridge Core i5 @ 2.6 GHz. That may help explain the difference. Best, Conrad ___ svn-src-all@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
Re: svn commit: r312910 - in head: . etc/etc.pc98 etc/rc.d lib/libsysdecode libexec release release/doc release/doc/en_US.ISO8859-1/hardware release/doc/en_US.ISO8859-1/readme release/doc/share/exampl
On Tue, Jan 31, 2017 at 04:36:04PM -0700, Warner Losh wrote: > ISA-only attachments would be a good place to start. My quick glance > shows this would be le(4), aha(4), cm(4), mse(4), and joy(4). There's > maybe half a dozen EISA-only drivers in the tree that I was planning > on retiring in 12, along with EISA. Likewise with the MCA attachments. I can join the list of people threatening to send ISA cards to anyone who objects. This should scare them into not objecting :-) (Yes, there is still at least one aha(4) card here in the house. Somewhere.) mcl ___ svn-src-all@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r313038 - stable/11/sys/net
Author: rpokala Date: Wed Feb 1 04:53:52 2017 New Revision: 313038 URL: https://svnweb.freebsd.org/changeset/base/313038 Log: MFC r312678: Eliminate misleading comments and dead code in lacp_port_create() Variables "fast" and "active" are both constant in lacp_port_create(), but comments mispleadingly suggest that "fast" can be changed via ioctl. The constant values control the value of "lp->lp_state", so it too is constant, and the code for assigning different value to it is essentially dead. Remove both "fast" and "active", and set "lp->lp_state" unconditionally; that gets rid of the dead code and misleading comments. Modified: stable/11/sys/net/ieee8023ad_lacp.c Directory Properties: stable/11/ (props changed) Modified: stable/11/sys/net/ieee8023ad_lacp.c == --- stable/11/sys/net/ieee8023ad_lacp.c Wed Feb 1 03:32:49 2017 (r313037) +++ stable/11/sys/net/ieee8023ad_lacp.c Wed Feb 1 04:53:52 2017 (r313038) @@ -526,9 +526,6 @@ lacp_port_create(struct lagg_port *lgp) struct ifmultiaddr *rifma = NULL; int error; - boolean_t active = TRUE; /* XXX should be configurable */ - boolean_t fast = FALSE; /* Configurable via ioctl */ - link_init_sdl(ifp, (struct sockaddr *)&sdl, IFT_ETHER); sdl.sdl_alen = ETHER_ADDR_LEN; @@ -557,9 +554,7 @@ lacp_port_create(struct lagg_port *lgp) lacp_fill_actorinfo(lp, &lp->lp_actor); lacp_fill_markerinfo(lp, &lp->lp_marker); - lp->lp_state = - (active ? LACP_STATE_ACTIVITY : 0) | - (fast ? LACP_STATE_TIMEOUT : 0); + lp->lp_state = LACP_STATE_ACTIVITY; lp->lp_aggregator = NULL; lacp_sm_rx_set_expired(lp); LACP_UNLOCK(lsc); ___ svn-src-all@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r313039 - stable/10/sys/net
Author: rpokala Date: Wed Feb 1 04:54:23 2017 New Revision: 313039 URL: https://svnweb.freebsd.org/changeset/base/313039 Log: MFC r312678: Eliminate misleading comments and dead code in lacp_port_create() Variables "fast" and "active" are both constant in lacp_port_create(), but comments mispleadingly suggest that "fast" can be changed via ioctl. The constant values control the value of "lp->lp_state", so it too is constant, and the code for assigning different value to it is essentially dead. Remove both "fast" and "active", and set "lp->lp_state" unconditionally; that gets rid of the dead code and misleading comments. Modified: stable/10/sys/net/ieee8023ad_lacp.c Directory Properties: stable/10/ (props changed) Modified: stable/10/sys/net/ieee8023ad_lacp.c == --- stable/10/sys/net/ieee8023ad_lacp.c Wed Feb 1 04:53:52 2017 (r313038) +++ stable/10/sys/net/ieee8023ad_lacp.c Wed Feb 1 04:54:23 2017 (r313039) @@ -523,9 +523,6 @@ lacp_port_create(struct lagg_port *lgp) struct ifmultiaddr *rifma = NULL; int error; - boolean_t active = TRUE; /* XXX should be configurable */ - boolean_t fast = FALSE; /* Configurable via ioctl */ - bzero((char *)&sdl, sizeof(sdl)); sdl.sdl_len = sizeof(sdl); sdl.sdl_family = AF_LINK; @@ -557,9 +554,7 @@ lacp_port_create(struct lagg_port *lgp) lacp_fill_actorinfo(lp, &lp->lp_actor); lacp_fill_markerinfo(lp, &lp->lp_marker); - lp->lp_state = - (active ? LACP_STATE_ACTIVITY : 0) | - (fast ? LACP_STATE_TIMEOUT : 0); + lp->lp_state = LACP_STATE_ACTIVITY; lp->lp_aggregator = NULL; lacp_sm_rx_set_expired(lp); LACP_UNLOCK(lsc); ___ svn-src-all@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r313040 - head/sys/mips/include
Author: kan
Date: Wed Feb 1 05:00:34 2017
New Revision: 313040
URL: https://svnweb.freebsd.org/changeset/base/313040
Log:
Add atomic_fcmpset_*() inlines for MIPS
atomic_fcmpset_*() is analogous to atomic_cmpset(), but saves off the
read value from the target memory location into the 'old' pointer.
Reviewed by: imp, brooks
Requested by: mjg
Differential Revision:https://reviews.freebsd.org/D9391
Modified:
head/sys/mips/include/atomic.h
Modified: head/sys/mips/include/atomic.h
==
--- head/sys/mips/include/atomic.h Wed Feb 1 04:54:23 2017
(r313039)
+++ head/sys/mips/include/atomic.h Wed Feb 1 05:00:34 2017
(r313040)
@@ -362,7 +362,7 @@ atomic_load_64(__volatile uint64_t *p, u
* zero if the compare failed, nonzero otherwise.
*/
static __inline uint32_t
-atomic_cmpset_32(__volatile uint32_t* p, uint32_t cmpval, uint32_t newval)
+atomic_cmpset_32(__volatile uint32_t *p, uint32_t cmpval, uint32_t newval)
{
uint32_t ret;
@@ -405,6 +405,46 @@ atomic_cmpset_rel_32(__volatile uint32_t
return (atomic_cmpset_32(p, cmpval, newval));
}
+static __inline uint32_t
+atomic_fcmpset_32(__volatile uint32_t *p, uint32_t *cmpval, uint32_t newval)
+{
+ uint32_t ret;
+
+ __asm __volatile (
+ "1:\n\t"
+ "ll %0, %1\n\t" /* load old value */
+ "bne%0, %4, 2f\n\t" /* compare */
+ "move %0, %3\n\t" /* value to store */
+ "sc %0, %1\n\t" /* attempt to store */
+ "beqz %0, 1b\n\t" /* if it failed, spin */
+ "j 3f\n\t"
+ "2:\n\t"
+ "sw %0, %2\n\t" /* save old value */
+ "li %0, 0\n\t"
+ "3:\n"
+ : "=&r" (ret), "+m" (*p), "=m" (*cmpval)
+ : "r" (newval), "r" (*cmpval)
+ : "memory");
+ return ret;
+}
+
+static __inline uint32_t
+atomic_fcmpset_acq_32(__volatile uint32_t *p, uint32_t *cmpval, uint32_t
newval)
+{
+ int retval;
+
+ retval = atomic_fcmpset_32(p, cmpval, newval);
+ mips_sync();
+ return (retval);
+}
+
+static __inline uint32_t
+atomic_fcmpset_rel_32(__volatile uint32_t *p, uint32_t *cmpval, uint32_t
newval)
+{
+ mips_sync();
+ return (atomic_fcmpset_32(p, cmpval, newval));
+}
+
/*
* Atomically add the value of v to the integer pointed to by p and return
* the previous value of *p.
@@ -431,7 +471,7 @@ atomic_fetchadd_32(__volatile uint32_t *
* zero if the compare failed, nonzero otherwise.
*/
static __inline uint64_t
-atomic_cmpset_64(__volatile uint64_t* p, uint64_t cmpval, uint64_t newval)
+atomic_cmpset_64(__volatile uint64_t *p, uint64_t cmpval, uint64_t newval)
{
uint64_t ret;
@@ -475,6 +515,47 @@ atomic_cmpset_rel_64(__volatile uint64_t
return (atomic_cmpset_64(p, cmpval, newval));
}
+static __inline uint32_t
+atomic_fcmpset_64(__volatile uint64_t *p, uint64_t *cmpval, uint64_t newval)
+{
+uint32_t ret;
+
+__asm __volatile (
+"1:\n\t"
+ "lld%0, %1\n\t" /* load old value */
+"bne %0, %4, 2f\n\t" /* compare */
+"move %0, %3\n\t" /* value to store */
+"scd %0, %1\n\t" /* attempt to store */
+"beqz %0, 1b\n\t" /* if it failed, spin */
+"j 3f\n\t"
+"2:\n\t"
+"sw%0, %2\n\t" /* save old value */
+"li%0, 0\n\t"
+"3:\n"
+: "=&r" (ret), "+m" (*p), "=m" (*cmpval)
+: "r" (newval), "r" (*cmpval)
+: "memory");
+
+ return ret;
+}
+
+static __inline uint64_t
+atomic_fcmpset_acq_64(__volatile uint64_t *p, uint64_t *cmpval, uint64_t
newval)
+{
+ int retval;
+
+ retval = atomic_fcmpset_64(p, cmpval, newval);
+ mips_sync();
+ return (retval);
+}
+
+static __inline uint64_t
+atomic_fcmpset_rel_64(__volatile uint64_t *p, uint64_t *cmpval, uint64_t
newval)
+{
+ mips_sync();
+ return (atomic_fcmpset_64(p, cmpval, newval));
+}
+
/*
* Atomically add the value of v to the integer pointed to by p and return
* the previous value of *p.
@@ -568,6 +649,9 @@ atomic_thread_fence_seq_cst(void)
#defineatomic_cmpset_int atomic_cmpset_32
#defineatomic_cmpset_acq_int atomic_cmpset_acq_32
#defineatomic_cmpset_rel_int atomic_cmpset_rel_32
+#defineatomic_fcmpset_int atomic_fcmpset_32
+#defineatomic_fcmpset_acq_int atomic_fcmpset_acq_32
+#defineatomic_fcmpset_rel_int atomic_fcmpset_rel_32
#defineatomic_load_acq_int atomic_load_acq_32
#defineatomic_store_rel_intatomic_st
svn commit: r313041 - head/gnu/usr.bin/cc/cc_tools
Author: emaste Date: Wed Feb 1 05:24:17 2017 New Revision: 313041 URL: https://svnweb.freebsd.org/changeset/base/313041 Log: Bump GCC FBSD_CC_VER for r312899 (-march=octeon+ support) Reported by: lidl MFC after:1 month MFC with: r312899 Sponsored by: The FreeBSD Foundation Modified: head/gnu/usr.bin/cc/cc_tools/freebsd-native.h Modified: head/gnu/usr.bin/cc/cc_tools/freebsd-native.h == --- head/gnu/usr.bin/cc/cc_tools/freebsd-native.h Wed Feb 1 05:00:34 2017(r313040) +++ head/gnu/usr.bin/cc/cc_tools/freebsd-native.h Wed Feb 1 05:24:17 2017(r313041) @@ -8,7 +8,7 @@ /* Fake out gcc/config/freebsd.h. */ #defineFBSD_MAJOR 12 -#defineFBSD_CC_VER 120 /* form like __FreeBSD_version */ +#defineFBSD_CC_VER 121 /* form like __FreeBSD_version */ #undef SYSTEM_INCLUDE_DIR /* We don't need one for now. */ #undef TOOL_INCLUDE_DIR/* We don't need one for now. */ ___ svn-src-all@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
