svn commit: r279947 - in head/share/doc: psd/title smm/title usd/title
Author: emaste Date: Fri Mar 13 07:29:49 2015 New Revision: 279947 URL: https://svnweb.freebsd.org/changeset/base/279947 Log: Remove UCB advertising clause per the letter dated July 22, 1999. Modified: head/share/doc/psd/title/Title head/share/doc/smm/title/Title head/share/doc/usd/title/Title Modified: head/share/doc/psd/title/Title == --- head/share/doc/psd/title/Title Fri Mar 13 04:26:48 2015 (r279946) +++ head/share/doc/psd/title/Title Fri Mar 13 07:29:49 2015 (r279947) @@ -9,11 +9,7 @@ .\" 2. Redistributions in binary form must reproduce the above copyright .\"notice, this list of conditions and the following disclaimer in the .\"documentation and/or other materials provided with the distribution. -.\" 3. All advertising materials mentioning features or use of this software -.\"must display the following acknowledgement: -.\"This product includes software developed by the University of -.\"California, Berkeley and its contributors. -.\" 4. Neither the name of the University nor the names of its contributors +.\" 3. Neither the name of the University nor the names of its contributors .\"may be used to endorse or promote products derived from this software .\"without specific prior written permission. .\" Modified: head/share/doc/smm/title/Title == --- head/share/doc/smm/title/Title Fri Mar 13 04:26:48 2015 (r279946) +++ head/share/doc/smm/title/Title Fri Mar 13 07:29:49 2015 (r279947) @@ -9,11 +9,7 @@ .\" 2. Redistributions in binary form must reproduce the above copyright .\"notice, this list of conditions and the following disclaimer in the .\"documentation and/or other materials provided with the distribution. -.\" 3. All advertising materials mentioning features or use of this software -.\"must display the following acknowledgement: -.\"This product includes software developed by the University of -.\"California, Berkeley and its contributors. -.\" 4. Neither the name of the University nor the names of its contributors +.\" 3. Neither the name of the University nor the names of its contributors .\"may be used to endorse or promote products derived from this software .\"without specific prior written permission. .\" Modified: head/share/doc/usd/title/Title == --- head/share/doc/usd/title/Title Fri Mar 13 04:26:48 2015 (r279946) +++ head/share/doc/usd/title/Title Fri Mar 13 07:29:49 2015 (r279947) @@ -9,11 +9,7 @@ .\" 2. Redistributions in binary form must reproduce the above copyright .\"notice, this list of conditions and the following disclaimer in the .\"documentation and/or other materials provided with the distribution. -.\" 3. All advertising materials mentioning features or use of this software -.\"must display the following acknowledgement: -.\"This product includes software developed by the University of -.\"California, Berkeley and its contributors. -.\" 4. Neither the name of the University nor the names of its contributors +.\" 3. Neither the name of the University nor the names of its contributors .\"may be used to endorse or promote products derived from this software .\"without specific prior written permission. .\" ___ svn-src-all@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
Re: svn commit: r279937 - in head/sys/powerpc: include powerpc
On Thu, Mar 12, 2015 at 02:46:53PM -0700, Nathan Whitehorn wrote:
>
> On 03/12/15 14:35, Konstantin Belousov wrote:
> > On Thu, Mar 12, 2015 at 02:29:43PM -0700, Nathan Whitehorn wrote:
> >> On 03/12/15 14:22, Konstantin Belousov wrote:
> >>> On Thu, Mar 12, 2015 at 09:15:39PM +, Nathan Whitehorn wrote:
> Author: nwhitehorn
> Date: Thu Mar 12 21:15:38 2015
> New Revision: 279937
> URL: https://svnweb.freebsd.org/changeset/base/279937
>
> Log:
> Provide VSX context in ucontext(3) API.
>
> Modified:
> head/sys/powerpc/include/ucontext.h
> head/sys/powerpc/powerpc/exec_machdep.c
>
> Modified: head/sys/powerpc/include/ucontext.h
> ==
> --- head/sys/powerpc/include/ucontext.h Thu Mar 12 20:14:48 2015
> (r279936)
> +++ head/sys/powerpc/include/ucontext.h Thu Mar 12 21:15:38 2015
> (r279937)
> @@ -46,6 +46,7 @@ typedef struct __mcontext {
> uint32_tmc_av[2];
> register_t mc_frame[42];
> uint64_tmc_fpreg[33];
> +uint64_tmc_vsxfpreg[32];/* low-order half of
> VSR0-31 */
> } mcontext_t __aligned(16);
>
> #if defined(_KERNEL) && defined(__powerpc64__)
> @@ -60,6 +61,7 @@ typedef struct __mcontext32 {
> uint32_tmc_av[2];
> uint32_tmc_frame[42];
> uint64_tmc_fpreg[33];
> +uint64_tmc_vsxfpreg[32];/* low-order half of
> VSR0-31 */
> } mcontext32_t __aligned(16);
> #endif
> >>> It looks as if you broken the ABI compatibility by the change. Am I
> >>> wrong ?
> >>>
> >> That is correct. It's a tier-2 platform and -CURRENT, so I'm not sure
> >> it's worth the compatibility shims. I'm happy to add them if you think
> >> otherwise.
> > You are main maintainer of PowerPC port, IMO, so it is your decision.
> >
> > Note that 'this is current' argument is not applicable, since the change
> > also breaks stable/* binaries.
> >
> > I do understand the argument of PowerPC being tier 2 architecture, but this
> > makes me sad. Anyway, it is yours. For x86, I have to introduce
> > getcontextx(3) mechanism.
> >
> This is a good point. I'll try to fix it. Is my understanding of how
> this works correct?
>
> 1. Provide a sysarch() for the extended FPU state.
> 2. Implement getcontextx() in the C library to fill extra properties if
> required.
> 3. Store state for signal trampoline in variable-sized stack area
4. Implement __getcontextx_size() and __fillcontextx2() for use
in the deferred signal delivery while libthr is in critical section,
see lib/libthr/thread/thr_sig.c:check_deferred_signal().
>
> Implementation of (2) seems to rely on having spare members in ucontext,
> which PowerPC unfortunately does not have. Is there a way around that?
Indeed, this is very unfortunate.
My concern is that typical application allocating ucontext_t on stack or
by mallocing it, would get silent memory corruption after the extension
of mcontext_t. It seems indeed that ABI breakage cannot be completely
avoided, but it could be significantly reduced IMO. Is it true that
mc_avec is only valid when _MC_AV_VALID bit is set ?
If yes, we can introduce another mcontext flag, say _MC_XSTATE_VALID,
which is mutually exclusive with the _MC_AC_VALID. The new flag
indicates that there is external data, and the data is pointed to by
some word placed in the previous mc_avec file, say mc_avec[0]. The
altivec registers file content is moved into that external data area as
well. Providing the external area size in mc_avec[1] allows to extend
that block in the future-compatible manner.
This way, applications which use ucontext_t and which are not aware about
VSX, get the expected behaviour, possibly without seeing altivec. I.e.
in typical case, we do not get random memory corruption.
Meantime, I have a question. I looked at the powerpc/include/ucontex.h
and tried to match it with the PowerISA specs 2.06 and 2.07. From what I
understand, mc_fpreg corresponds to the floating-point registers file,
mc_avec and mc_av to the 'Vector Facility Registers'. But I fail to see
what would match the
+ uint64_tmc_vsxfpreg[32];/* low-order half of VSR0-31 */
file. The 7.2.1 Vector-Scalar Registers says 'Sixty-four 128-bit VSRs
are provided'.
___
svn-src-all@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/svn-src-all
To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r279948 - in head: sbin/ipfw sys/netinet sys/netpfil/ipfw
Author: ae
Date: Fri Mar 13 09:03:25 2015
New Revision: 279948
URL: https://svnweb.freebsd.org/changeset/base/279948
Log:
Fix `ipfw fwd tablearg'. Use dedicated field nh4 in struct table_value
to obtain IPv4 next hop address in tablearg case.
Add `fwd tablearg' support for IPv6. ipfw(8) uses INADDR_ANY as next hop
address in O_FORWARD_IP opcode for specifying tablearg case. For IPv6 we
still use this opcode, but when packet identified as IPv6 packet, we
obtain next hop address from dedicated field nh6 in struct table_value.
Replace hopstore field in struct ip_fw_args with anonymous union and add
hopstore6 field. Use this field to copy tablearg value for IPv6.
Replace spare1 field in struct table_value with zoneid. Use it to keep
scope zone id for link-local IPv6 addresses. Since spare1 was used
internally, replace spare0 array with two variables spare0 and spare1.
Use getaddrinfo(3)/getnameinfo(3) functions for parsing and formatting
IPv6 addresses in table_value. Use zoneid field in struct table_value
to store sin6_scope_id value.
Since the kernel still uses embedded scope zone id to represent
link-local addresses, convert next_hop6 address into this form before
return from pfil processing. This also fixes in6_localip() check
for link-local addresses.
Differential Revision:https://reviews.freebsd.org/D2015
Obtained from:Yandex LLC
Sponsored by: Yandex LLC
Modified:
head/sbin/ipfw/ipfw.8
head/sbin/ipfw/tables.c
head/sys/netinet/ip_fw.h
head/sys/netpfil/ipfw/ip_fw2.c
head/sys/netpfil/ipfw/ip_fw_pfil.c
head/sys/netpfil/ipfw/ip_fw_private.h
head/sys/netpfil/ipfw/ip_fw_table_value.c
Modified: head/sbin/ipfw/ipfw.8
==
--- head/sbin/ipfw/ipfw.8 Fri Mar 13 07:29:49 2015(r279947)
+++ head/sbin/ipfw/ipfw.8 Fri Mar 13 09:03:25 2015(r279948)
@@ -1,7 +1,7 @@
.\"
.\" $FreeBSD$
.\"
-.Dd Aug 13, 2014
+.Dd March 13, 2015
.Dt IPFW 8
.Os
.Sh NAME
@@ -2078,6 +2078,8 @@ hook number to move packet to.
maximum number of connections.
.It Cm ipv4
IPv4 nexthop to fwd packets to.
+.It Cm ipv6
+IPv6 nexthop to fwd packets to.
.El
.Pp
The
Modified: head/sbin/ipfw/tables.c
==
--- head/sbin/ipfw/tables.c Fri Mar 13 07:29:49 2015(r279947)
+++ head/sbin/ipfw/tables.c Fri Mar 13 09:03:25 2015(r279948)
@@ -35,6 +35,7 @@
#include
#include
#include
+#include
#include "ipfw2.h"
@@ -1384,6 +1385,7 @@ static void
tentry_fill_value(ipfw_obj_header *oh, ipfw_obj_tentry *tent, char *arg,
uint8_t type, uint32_t vmask)
{
+ struct addrinfo hints, *res;
uint32_t a4, flag, val, vm;
ipfw_table_value *v;
uint32_t i;
@@ -1494,9 +1496,19 @@ tentry_fill_value(ipfw_obj_header *oh, i
}
break;
case IPFW_VTYPE_NH6:
- if (strchr(n, ':') != NULL &&
- inet_pton(AF_INET6, n, &v->nh6) == 1)
- break;
+ if (strchr(n, ':') != NULL) {
+ memset(&hints, 0, sizeof(hints));
+ hints.ai_family = AF_INET6;
+ hints.ai_flags = AI_NUMERICHOST;
+ if (getaddrinfo(n, NULL, &hints, &res) == 0) {
+ v->nh6 = ((struct sockaddr_in6 *)
+ res->ai_addr)->sin6_addr;
+ v->zoneid = ((struct sockaddr_in6 *)
+ res->ai_addr)->sin6_scope_id;
+ freeaddrinfo(res);
+ break;
+ }
+ }
etype = "ipv6";
break;
}
@@ -1643,10 +1655,11 @@ static void
table_show_value(char *buf, size_t bufsize, ipfw_table_value *v,
uint32_t vmask, int print_ip)
{
+ char abuf[INET6_ADDRSTRLEN + IF_NAMESIZE + 2];
+ struct sockaddr_in6 sa6;
uint32_t flag, i, l;
size_t sz;
struct in_addr a4;
- char abuf[INET6_ADDRSTRLEN];
sz = bufsize;
@@ -1702,8 +1715,15 @@ table_show_value(char *buf, size_t bufsi
l = snprintf(buf, sz, "%d,", v->dscp);
break;
case IPFW_VTYPE_NH6:
- inet_ntop(AF_INET6, &v->nh6, abuf, sizeof(abuf));
- l = snprintf(buf, sz, "%s,", abuf);
+ sa6.sin6_family = AF_INET6;
+ sa6.sin6_len = sizeof(sa6);
+ sa6.sin6_addr = v->nh6;
+ sa6.sin6_port = 0;
+ sa6.sin6_scope_id = v-
svn commit: r279949 - in head: lib/libstand sys/boot/amd64 sys/boot/efi/libefi sys/boot/ficl
Author: jhb
Date: Fri Mar 13 09:38:16 2015
New Revision: 279949
URL: https://svnweb.freebsd.org/changeset/base/279949
Log:
The System V ABI for amd64 allows functions to use space in a 128 byte
redzone below the stack pointer for scratch space and requires
interrupt and signal frames to avoid overwriting it. However, EFI uses
the Windows ABI which does not support this. As a result, interrupt
handlers in EFI push their interrupt frames directly on top of the
stack pointer. If the compiler used the red zone in a function in the
EFI loader, then a device interrupt that occurred while that function
was running could trash its local variables. In practice this happens
fairly reliable when using gzipfs as an interrupt during decompression
can trash the local variables in the inflate_table() function
resulting in corrupted output or hangs.
Fix this by disabling the redzone for amd64 EFI binaries. This
requires building not only the loader but any libraries used by the
loader without redzone support.
Thanks to Jilles for pointing me at the redzone once I found the stack
corruption.
Differential Revision:https://reviews.freebsd.org/D2054
Reviewed by: imp
MFC after:2 weeks
Sponsored by: Cisco Systems, Inc.
Modified:
head/lib/libstand/Makefile
head/sys/boot/amd64/Makefile.inc
head/sys/boot/efi/libefi/Makefile
head/sys/boot/ficl/Makefile
Modified: head/lib/libstand/Makefile
==
--- head/lib/libstand/Makefile Fri Mar 13 09:03:25 2015(r279948)
+++ head/lib/libstand/Makefile Fri Mar 13 09:38:16 2015(r279949)
@@ -28,7 +28,7 @@ CFLAGS+= -mno-mmx -mno-3dnow -mno-sse -m
CFLAGS.gcc+= -mpreferred-stack-boundary=2
.endif
.if ${MACHINE_CPUARCH} == "amd64"
-CFLAGS+= -fPIC
+CFLAGS+= -fPIC -mno-red-zone
.endif
.if ${MACHINE} == "pc98"
CFLAGS+= -Os
Modified: head/sys/boot/amd64/Makefile.inc
==
--- head/sys/boot/amd64/Makefile.incFri Mar 13 09:03:25 2015
(r279948)
+++ head/sys/boot/amd64/Makefile.incFri Mar 13 09:38:16 2015
(r279949)
@@ -5,7 +5,7 @@
BINDIR?= /boot
# See conf/kern.mk for the correct set of these
-CFLAGS+= -ffreestanding
+CFLAGS+= -ffreestanding -mno-red-zone
CFLAGS+= -mno-mmx -mno-sse -mno-aes -mno-avx -msoft-float
LDFLAGS+= -nostdlib
Modified: head/sys/boot/efi/libefi/Makefile
==
--- head/sys/boot/efi/libefi/Makefile Fri Mar 13 09:03:25 2015
(r279948)
+++ head/sys/boot/efi/libefi/Makefile Fri Mar 13 09:38:16 2015
(r279949)
@@ -9,7 +9,7 @@ SRCS= delay.c efi_console.c efinet.c efi
SRCS+= nullconsole.c comconsole.c
.if ${MACHINE_ARCH} == "amd64"
-CFLAGS+= -fPIC
+CFLAGS+= -fPIC -mno-red-zone
.endif
CFLAGS+= -I${.CURDIR}/../include
CFLAGS+= -I${.CURDIR}/../include/${MACHINE_CPUARCH}
Modified: head/sys/boot/ficl/Makefile
==
--- head/sys/boot/ficl/Makefile Fri Mar 13 09:03:25 2015(r279948)
+++ head/sys/boot/ficl/Makefile Fri Mar 13 09:38:16 2015(r279949)
@@ -21,6 +21,9 @@ CFLAGS+= -ffreestanding
CFLAGS+= -march=i386
CFLAGS.gcc+= -mpreferred-stack-boundary=2
.endif
+.if ${MACHINE_CPUARCH} == "amd64"
+CFLAGS+= -mno-red-zone
+.endif
.if ${MACHINE_CPUARCH} == "i386" || ${MACHINE_CPUARCH} == "amd64"
CFLAGS+= -mno-mmx -mno-3dnow -mno-sse -mno-sse2 -mno-sse3 -msoft-float
.endif
___
svn-src-all@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/svn-src-all
To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r279950 - in head/sys/boot: amd64/efi efi/libefi
Author: jhb
Date: Fri Mar 13 09:41:27 2015
New Revision: 279950
URL: https://svnweb.freebsd.org/changeset/base/279950
Log:
Enable bzipfs support in the EFI loader.
- Add bzipfs to the list of supported filesystems in the EFI loader.
- Increase the heap size allocated for the EFI loader from 2MB to 3MB.
Differential Revision:https://reviews.freebsd.org/D2053
Reviewed by: benno, emaste, imp
MFC after:2 weeks
Sponsored by: Cisco Systems, Inc.
Modified:
head/sys/boot/amd64/efi/conf.c
head/sys/boot/efi/libefi/libefi.c
Modified: head/sys/boot/amd64/efi/conf.c
==
--- head/sys/boot/amd64/efi/conf.c Fri Mar 13 09:38:16 2015
(r279949)
+++ head/sys/boot/amd64/efi/conf.c Fri Mar 13 09:41:27 2015
(r279950)
@@ -44,6 +44,7 @@ struct fs_ops *file_system[] = {
&cd9660_fsops,
&nfs_fsops,
&gzipfs_fsops,
+ &bzipfs_fsops,
NULL
};
Modified: head/sys/boot/efi/libefi/libefi.c
==
--- head/sys/boot/efi/libefi/libefi.c Fri Mar 13 09:38:16 2015
(r279949)
+++ head/sys/boot/efi/libefi/libefi.c Fri Mar 13 09:41:27 2015
(r279950)
@@ -102,7 +102,7 @@ efi_main(EFI_HANDLE image_handle, EFI_SY
(void)console_control->SetMode(console_control,
EfiConsoleControlScreenText);
- heapsize = 2 * 1024 * 1024;
+ heapsize = 3 * 1024 * 1024;
status = BS->AllocatePages(AllocateAnyPages, EfiLoaderData,
EFI_SIZE_TO_PAGES(heapsize), &heap);
if (status != EFI_SUCCESS)
___
svn-src-all@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/svn-src-all
To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r279951 - head/sbin/ifconfig
Author: jhb Date: Fri Mar 13 09:45:06 2015 New Revision: 279951 URL: https://svnweb.freebsd.org/changeset/base/279951 Log: Simplify string mangling in ifmaybeload(). - Use strlcpy() instead of strcpy(). - Use strlcat() instead of a strlcpy() with a magic number subtracted from the length. - Replace strncmp(..., strlen(foo) + 1) with strcmp(...). Differential Revision:https://reviews.freebsd.org/D1814 Reviewed by: rpaulo MFC after:2 weeks Modified: head/sbin/ifconfig/ifconfig.c Modified: head/sbin/ifconfig/ifconfig.c == --- head/sbin/ifconfig/ifconfig.c Fri Mar 13 09:41:27 2015 (r279950) +++ head/sbin/ifconfig/ifconfig.c Fri Mar 13 09:45:06 2015 (r279951) @@ -1280,9 +1280,8 @@ ifmaybeload(const char *name) } /* turn interface and unit into module name */ - strcpy(ifkind, "if_"); - strlcpy(ifkind + MOD_PREFIX_LEN, ifname, - sizeof(ifkind) - MOD_PREFIX_LEN); + strlcpy(ifkind, "if_", sizeof(ifkind)); + strlcat(ifkind, ifname, sizeof(ifkind)); /* scan files in kernel */ mstat.version = sizeof(struct module_stat); @@ -1299,8 +1298,8 @@ ifmaybeload(const char *name) cp = mstat.name; } /* already loaded? */ - if (strncmp(ifname, cp, strlen(ifname) + 1) == 0 || - strncmp(ifkind, cp, strlen(ifkind) + 1) == 0) + if (strcmp(ifname, cp) == 0 || + strcmp(ifkind, cp) == 0) return; } } ___ svn-src-all@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r279952 - in head: etc/defaults share/man/man5
Author: jhb Date: Fri Mar 13 09:50:29 2015 New Revision: 279952 URL: https://svnweb.freebsd.org/changeset/base/279952 Log: - Align comment for df flags variable in periodic.conf. - Note default value of df flags variable in periodoc.conf(5). MFC after:1 week Modified: head/etc/defaults/periodic.conf head/share/man/man5/periodic.conf.5 Modified: head/etc/defaults/periodic.conf == --- head/etc/defaults/periodic.conf Fri Mar 13 09:45:06 2015 (r279951) +++ head/etc/defaults/periodic.conf Fri Mar 13 09:50:29 2015 (r279952) @@ -89,7 +89,7 @@ daily_news_expire_enable="YES" # Run # 400.status-disks daily_status_disks_enable="YES"# Check disk status -daily_status_disks_df_flags="-l -h"# df(1) flags for check +daily_status_disks_df_flags="-l -h"# df(1) flags for check # 401.status-graid daily_status_graid_enable="NO" # Check graid(8) Modified: head/share/man/man5/periodic.conf.5 == --- head/share/man/man5/periodic.conf.5 Fri Mar 13 09:45:06 2015 (r279951) +++ head/share/man/man5/periodic.conf.5 Fri Mar 13 09:50:29 2015 (r279952) @@ -25,7 +25,7 @@ .\" .\" $FreeBSD$ .\" -.Dd January 15, 2015 +.Dd March 13, 2015 .Dt PERIODIC.CONF 5 .Os .Sh NAME @@ -335,6 +335,8 @@ utility when .Va daily_status_disks_enable is set to .Dq Li YES . +The default is +.Fl l Fl h . .It Va daily_status_zfs_enable .Pq Vt bool Set to ___ svn-src-all@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
Re: svn commit: r279932 - head/sys/vm
On Thursday, March 12, 2015 05:24:51 PM Ian Lepore wrote: > On Thu, 2015-03-12 at 17:02 -0400, Ryan Stone wrote: > > On Thu, Mar 12, 2015 at 2:06 PM, Ian Lepore wrote: > > > > > Nullterminate strings returned via sysctl. > > > > > > PR: 195668 > > > > > > > To quote the manpage: > > > > > The *sbuf* family of functions allows one to safely > > > allocate, construct and release bounded null-terminated > > > strings in kernel space. > > > > IMO the sbuf API is broken if we have to explicitly null-terminate the > > string ourselves. > > If we want the nullterm to be counted in the length of data in the > buffer (and thus get transmitted back across the syscall boundary) we > need to put an explicit counted nullterm byte into the buffer. > > I had started down the path of making that (counting the nullterm byte > as part of the data in the buffer) a feature of sbuf that you could set > with a flag, but then whoever added sbuf_new_for_sysctl() didn't > propagate the flags field through the new function and I decided to not > go off into the weeds making a new flavor of that takes flags. One suggestion would be to consider using '\0' for a nul character instead of a bare 0. To me that communicates the intention more clearly to the reader. (One of the things I did not like about C++ < C++11 was the use of 0 for NULL. I much prefer nullptr and NULL in C over bare 0's for pointers for similar clarity reasons.) -- John Baldwin ___ svn-src-all@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
Re: svn commit: r279949 - in head: lib/libstand sys/boot/amd64 sys/boot/efi/libefi sys/boot/ficl
On Friday, March 13, 2015 09:38:16 AM John Baldwin wrote: > Author: jhb > Date: Fri Mar 13 09:38:16 2015 > New Revision: 279949 > URL: https://svnweb.freebsd.org/changeset/base/279949 > > Log: > The System V ABI for amd64 allows functions to use space in a 128 byte > redzone below the stack pointer for scratch space and requires > interrupt and signal frames to avoid overwriting it. However, EFI uses > the Windows ABI which does not support this. As a result, interrupt > handlers in EFI push their interrupt frames directly on top of the > stack pointer. If the compiler used the red zone in a function in the > EFI loader, then a device interrupt that occurred while that function > was running could trash its local variables. In practice this happens > fairly reliable when using gzipfs as an interrupt during decompression > can trash the local variables in the inflate_table() function > resulting in corrupted output or hangs. > > Fix this by disabling the redzone for amd64 EFI binaries. This > requires building not only the loader but any libraries used by the > loader without redzone support. > > Thanks to Jilles for pointing me at the redzone once I found the stack > corruption. Note that without this, gzipfs was effectively useless in the EFI loader for any non-trivial cases (it was even useless in my trivial cases during testing, but there might be some trivial case (zero-byte file or some such) that can be decompressed quickly enough to not trigger a device interrupt). Also, while the redzone is cute and all, it seems to me that it is very impractical for any ring 0 context. Specifically, the CPU doesn't know anything about it, and any interrupt or exception in ring 0 is going to push the hardware-defined bits of the frame (%rip, etc.) on top of the stack pointer overwriting part of the redzone. (Our kernel is built with -mno-red-zone as well.) You could perhaps always use a dedicated stack for all interrupts and faults via IST, but any nesting there would result in reusing a stack if you aren't careful (e.g. a fault in interrupt context). I suppose it might work to have one IST entry for device interrupts, a second for most faults / exceptions, and a third for double faults (double faults generally need a dedicated stack regardless). Of course, all these dedicated stacks would have to be per-CPU (as the double fault stack is) and any fault or interrupt would have to implicitly pin the thread for the duration (probably not a big deal). -- John Baldwin ___ svn-src-all@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r279953 - head/usr.sbin/autofs
Author: trasz
Date: Fri Mar 13 11:26:02 2015
New Revision: 279953
URL: https://svnweb.freebsd.org/changeset/base/279953
Log:
Rework the concat() algorithm to be correct in all cases.
MFC after:1 month
Sponsored by: The FreeBSD Foundation
Modified:
head/usr.sbin/autofs/common.c
Modified: head/usr.sbin/autofs/common.c
==
--- head/usr.sbin/autofs/common.c Fri Mar 13 09:50:29 2015
(r279952)
+++ head/usr.sbin/autofs/common.c Fri Mar 13 11:26:02 2015
(r279953)
@@ -92,6 +92,7 @@ char *
concat(const char *s1, char separator, const char *s2)
{
char *result;
+ char s1last, s2first;
int ret;
if (s1 == NULL)
@@ -99,14 +100,22 @@ concat(const char *s1, char separator, c
if (s2 == NULL)
s2 = "";
- /*
-* If s2 starts with separator - skip it; otherwise concatenating
-* "/" and "/foo" would end up returning "//foo".
-*/
- if (s2[0] == separator)
- s2++;
+ if (s1[0] == '\0')
+ s1last = '\0';
+ else
+ s1last = s1[strlen(s1) - 1];
- if (s1[0] == '\0' || s2[0] == '\0' || s1[strlen(s1) - 1] == separator) {
+ s2first = s2[0];
+
+ if (s1last == separator && s2first == separator) {
+ /*
+* If s1 ends with the separator and s2 begins with
+* it - skip the latter; otherwise concatenating "/"
+* and "/foo" would end up returning "//foo".
+*/
+ ret = asprintf(&result, "%s%s", s1, s2 + 1);
+ } else if (s1last == separator || s2first == separator ||
+ s1[0] == '\0' || s2[0] == '\0') {
ret = asprintf(&result, "%s%s", s1, s2);
} else {
ret = asprintf(&result, "%s%c%s", s1, separator, s2);
___
svn-src-all@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/svn-src-all
To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r279954 - head/usr.sbin/autofs
Author: trasz
Date: Fri Mar 13 12:27:59 2015
New Revision: 279954
URL: https://svnweb.freebsd.org/changeset/base/279954
Log:
Get executable direct maps to work.
MFC after:1 month
Sponsored by: The FreeBSD Foundation
Modified:
head/usr.sbin/autofs/automountd.c
Modified: head/usr.sbin/autofs/automountd.c
==
--- head/usr.sbin/autofs/automountd.c Fri Mar 13 11:26:02 2015
(r279953)
+++ head/usr.sbin/autofs/automountd.c Fri Mar 13 12:27:59 2015
(r279954)
@@ -177,7 +177,7 @@ handle_request(const struct autofs_daemo
const char *map;
struct node *root, *parent, *node;
FILE *f;
- char *options, *fstype, *nobrowse, *retrycnt, *tmp;
+ char *key, *options, *fstype, *nobrowse, *retrycnt, *tmp;
int error;
bool wildcards;
@@ -199,11 +199,25 @@ handle_request(const struct autofs_daemo
map = adr->adr_from + 4; /* 4 for strlen("map "); */
root = node_new_root();
if (adr->adr_prefix[0] == '\0' || strcmp(adr->adr_prefix, "/") == 0) {
+ /*
+* Direct map. autofs(4) doesn't have a way to determine
+* correct map key, but since it's a direct map, we can just
+* use adr_path instead.
+*/
parent = root;
+ key = checked_strdup(adr->adr_path);
} else {
+ /*
+* Indirect map.
+*/
parent = node_new_map(root, checked_strdup(adr->adr_prefix),
NULL, checked_strdup(map),
checked_strdup("[kernel request]"), lineno);
+
+ if (adr->adr_key[0] == '\0')
+ key = NULL;
+ else
+ key = checked_strdup(adr->adr_key);
}
/*
@@ -213,8 +227,7 @@ handle_request(const struct autofs_daemo
* needs to be done for maps with wildcard entries, but also
* for special and executable maps.
*/
- parse_map(parent, map, adr->adr_key[0] != '\0' ? adr->adr_key : NULL,
- &wildcards);
+ parse_map(parent, map, key, &wildcards);
if (!wildcards)
wildcards = node_has_wildcards(parent);
if (wildcards)
@@ -222,8 +235,8 @@ handle_request(const struct autofs_daemo
else
log_debugx("map does not contain wildcard entries");
- if (adr->adr_key[0] != '\0')
- node_expand_wildcard(root, adr->adr_key);
+ if (key != NULL)
+ node_expand_wildcard(root, key);
node = node_find(root, adr->adr_path);
if (node == NULL) {
@@ -248,7 +261,7 @@ handle_request(const struct autofs_daemo
node->n_config_file, node->n_config_line);
nobrowse = pick_option("nobrowse", &options);
- if (nobrowse != NULL && adr->adr_key[0] == '\0') {
+ if (nobrowse != NULL && key == NULL) {
log_debugx("skipping map %s due to \"nobrowse\" "
"option; exiting", map);
done(0, true);
@@ -265,12 +278,12 @@ handle_request(const struct autofs_daemo
*/
create_subtree(node, incomplete_hierarchy);
- if (incomplete_hierarchy && adr->adr_key[0] != '\0') {
+ if (incomplete_hierarchy && key != NULL) {
/*
* We still need to create the single subdirectory
* user is trying to access.
*/
- tmp = concat(adr->adr_path, '/', adr->adr_key);
+ tmp = concat(adr->adr_path, '/', key);
node = node_find(root, tmp);
if (node != NULL)
create_subtree(node, false);
@@ -288,8 +301,8 @@ handle_request(const struct autofs_daemo
log_debugx("found node defined at %s:%d; it is a mountpoint",
node->n_config_file, node->n_config_line);
- node_expand_ampersand(node,
- adr->adr_key[0] != '\0' ? adr->adr_key : NULL);
+ if (key != NULL)
+ node_expand_ampersand(node, key);
error = node_expand_defined(node);
if (error != 0) {
log_errx(1, "variable expansion failed for %s; "
___
svn-src-all@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/svn-src-all
To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r279955 - in head: etc etc/autofs usr.sbin/autofs
Author: trasz
Date: Fri Mar 13 12:44:46 2015
New Revision: 279955
URL: https://svnweb.freebsd.org/changeset/base/279955
Log:
Add -noauto autofs map, for automatic handling of fstab entries
marked "noauto".
MFC after:1 month
Sponsored by: The FreeBSD Foundation
Added:
head/etc/autofs/special_noauto (contents, props changed)
Modified:
head/etc/auto_master
head/etc/autofs/Makefile
head/usr.sbin/autofs/auto_master.5
Modified: head/etc/auto_master
==
--- head/etc/auto_masterFri Mar 13 12:27:59 2015(r279954)
+++ head/etc/auto_masterFri Mar 13 12:44:46 2015(r279955)
@@ -6,3 +6,4 @@
# When using the -media special map, make sure to edit devd.conf(5)
# to move the call to "automount -c" out of the comments section.
#/media-media -nosuid
+#/--noauto
Modified: head/etc/autofs/Makefile
==
--- head/etc/autofs/MakefileFri Mar 13 12:27:59 2015(r279954)
+++ head/etc/autofs/MakefileFri Mar 13 12:44:46 2015(r279955)
@@ -1,6 +1,6 @@
# $FreeBSD$
-FILES= include_ldap special_hosts special_media special_null
+FILES= include_ldap special_hosts special_media special_noauto
special_null
NO_OBJ=
FILESDIR= /etc/autofs
Added: head/etc/autofs/special_noauto
==
--- /dev/null 00:00:00 1970 (empty, because file is newly added)
+++ head/etc/autofs/special_noauto Fri Mar 13 12:44:46 2015
(r279955)
@@ -0,0 +1,29 @@
+#!/bin/sh
+#
+# $FreeBSD$
+#
+
+print_available() {
+ sed 's/#.*//' /etc/fstab | awk '$4 ~ /noauto/ { print $2 }'
+}
+
+print_one() {
+ local _mntpoint
+
+ _mntpoint="${1%/}"
+
+ sed 's/#.*//' /etc/fstab | awk '
+ $2 == "'"${_mntpoint}"'" && $4 ~ /noauto/ {
+ if ($1 ~ /:/) { dev=$1 } else { dev=":"$1 }
+ print "-fstype=" $3 "," $4, dev
+ }'
+}
+
+if [ $# -eq 0 ]; then
+ print_available
+ exit 0
+fi
+
+print_one "$1"
+exit 0
+
Modified: head/usr.sbin/autofs/auto_master.5
==
--- head/usr.sbin/autofs/auto_master.5 Fri Mar 13 12:27:59 2015
(r279954)
+++ head/usr.sbin/autofs/auto_master.5 Fri Mar 13 12:44:46 2015
(r279955)
@@ -27,7 +27,7 @@
.\"
.\" $FreeBSD$
.\"
-.Dd January 9, 2015
+.Dd March 13, 2015
.Dt AUTO_MASTER 5
.Os
.Sh NAME
@@ -260,6 +260,11 @@ when they are automatically created.
.It Li -media
Query devices that are not yet mounted, but contain valid filesystems.
Generally used to access files on removable media.
+.It Li -noauto
+Mount filesystems configured in
+.Xr fstab 5
+as "noauto".
+This needs to be set up as a direct map.
.It Li -null
Prevent
.Xr automountd 8
___
svn-src-all@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/svn-src-all
To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r279956 - head/sbin/ipfw
Author: ae
Date: Fri Mar 13 13:46:50 2015
New Revision: 279956
URL: https://svnweb.freebsd.org/changeset/base/279956
Log:
Properly initialize scope zone id when next hop address stored
directly in the O_FORWARD_IP6 opcode. Use getnameinfo(3) to formatting
the IPv6 addresses of such opcodes.
Obtained from:Yandex LLC
Sponsored by: Yandex LLC
Modified:
head/sbin/ipfw/ipfw2.c
Modified: head/sbin/ipfw/ipfw2.c
==
--- head/sbin/ipfw/ipfw2.c Fri Mar 13 12:44:46 2015(r279955)
+++ head/sbin/ipfw/ipfw2.c Fri Mar 13 13:46:50 2015(r279956)
@@ -1525,11 +1525,14 @@ show_static_rule(struct cmdline_opts *co
case O_FORWARD_IP6:
{
- char buf[4 + INET6_ADDRSTRLEN + 1];
+ char buf[INET6_ADDRSTRLEN + IF_NAMESIZE + 2];
ipfw_insn_sa6 *s = (ipfw_insn_sa6 *)cmd;
- bprintf(bp, "fwd %s", inet_ntop(AF_INET6,
- &s->sa.sin6_addr, buf, sizeof(buf)));
+ bprintf(bp, "fwd ");
+ if (getnameinfo((const struct sockaddr *)&s->sa,
+ sizeof(struct sockaddr_in6), buf, sizeof(buf),
+ NULL, 0, NI_NUMERICHOST) == 0)
+ bprintf(bp, "%s", buf);
if (s->sa.sin6_port)
bprintf(bp, ",%d", s->sa.sin6_port);
}
@@ -3741,8 +3744,8 @@ chkarg:
p->sa.sin6_family = AF_INET6;
p->sa.sin6_port = port_number;
p->sa.sin6_flowinfo = 0;
- p->sa.sin6_scope_id = 0;
- /* No table support for v6 yet. */
+ p->sa.sin6_scope_id =
+ ((struct sockaddr_in6 *)&result)->sin6_scope_id;
bcopy(&((struct sockaddr_in6*)&result)->sin6_addr,
&p->sa.sin6_addr, sizeof(p->sa.sin6_addr));
} else {
___
svn-src-all@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/svn-src-all
To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
Re: svn commit: r279937 - in head/sys/powerpc: include powerpc
On 03/13/15 01:47, Konstantin Belousov wrote:
On Thu, Mar 12, 2015 at 02:46:53PM -0700, Nathan Whitehorn wrote:
On 03/12/15 14:35, Konstantin Belousov wrote:
On Thu, Mar 12, 2015 at 02:29:43PM -0700, Nathan Whitehorn wrote:
On 03/12/15 14:22, Konstantin Belousov wrote:
On Thu, Mar 12, 2015 at 09:15:39PM +, Nathan Whitehorn wrote:
Author: nwhitehorn
Date: Thu Mar 12 21:15:38 2015
New Revision: 279937
URL: https://svnweb.freebsd.org/changeset/base/279937
Log:
Provide VSX context in ucontext(3) API.
Modified:
head/sys/powerpc/include/ucontext.h
head/sys/powerpc/powerpc/exec_machdep.c
Modified: head/sys/powerpc/include/ucontext.h
==
--- head/sys/powerpc/include/ucontext.h Thu Mar 12 20:14:48 2015
(r279936)
+++ head/sys/powerpc/include/ucontext.h Thu Mar 12 21:15:38 2015
(r279937)
@@ -46,6 +46,7 @@ typedef struct __mcontext {
uint32_tmc_av[2];
register_t mc_frame[42];
uint64_tmc_fpreg[33];
+ uint64_tmc_vsxfpreg[32];/* low-order half of VSR0-31 */
} mcontext_t __aligned(16);
#if defined(_KERNEL) && defined(__powerpc64__)
@@ -60,6 +61,7 @@ typedef struct __mcontext32 {
uint32_tmc_av[2];
uint32_tmc_frame[42];
uint64_tmc_fpreg[33];
+ uint64_tmc_vsxfpreg[32];/* low-order half of VSR0-31 */
} mcontext32_t __aligned(16);
#endif
It looks as if you broken the ABI compatibility by the change. Am I wrong ?
That is correct. It's a tier-2 platform and -CURRENT, so I'm not sure
it's worth the compatibility shims. I'm happy to add them if you think
otherwise.
You are main maintainer of PowerPC port, IMO, so it is your decision.
Note that 'this is current' argument is not applicable, since the change
also breaks stable/* binaries.
I do understand the argument of PowerPC being tier 2 architecture, but this
makes me sad. Anyway, it is yours. For x86, I have to introduce
getcontextx(3) mechanism.
This is a good point. I'll try to fix it. Is my understanding of how
this works correct?
1. Provide a sysarch() for the extended FPU state.
2. Implement getcontextx() in the C library to fill extra properties if
required.
3. Store state for signal trampoline in variable-sized stack area
4. Implement __getcontextx_size() and __fillcontextx2() for use
in the deferred signal delivery while libthr is in critical section,
see lib/libthr/thread/thr_sig.c:check_deferred_signal().
OK.
Implementation of (2) seems to rely on having spare members in ucontext,
which PowerPC unfortunately does not have. Is there a way around that?
Indeed, this is very unfortunate.
My concern is that typical application allocating ucontext_t on stack or
by mallocing it, would get silent memory corruption after the extension
of mcontext_t. It seems indeed that ABI breakage cannot be completely
avoided, but it could be significantly reduced IMO. Is it true that
mc_avec is only valid when _MC_AV_VALID bit is set ?
That is correct.
If yes, we can introduce another mcontext flag, say _MC_XSTATE_VALID,
which is mutually exclusive with the _MC_AC_VALID. The new flag
indicates that there is external data, and the data is pointed to by
some word placed in the previous mc_avec file, say mc_avec[0]. The
altivec registers file content is moved into that external data area as
well. Providing the external area size in mc_avec[1] allows to extend
that block in the future-compatible manner.
This way, applications which use ucontext_t and which are not aware about
VSX, get the expected behaviour, possibly without seeing altivec. I.e.
in typical case, we do not get random memory corruption.
OK, I guess that's reasonable. I'll think about this some and try to get
some code written in the next few days.
Meantime, I have a question. I looked at the powerpc/include/ucontex.h
and tried to match it with the PowerISA specs 2.06 and 2.07. From what I
understand, mc_fpreg corresponds to the floating-point registers file,
mc_avec and mc_av to the 'Vector Facility Registers'. But I fail to see
what would match the
+ uint64_tmc_vsxfpreg[32];/* low-order half of VSR0-31 */
file. The 7.2.1 Vector-Scalar Registers says 'Sixty-four 128-bit VSRs
are provided'.
The 64 128-bit registers are a superset of the existing registers.
Registers 33-64 are just the Altivec registers, with no changes.
Registers 1-32 are the normal floating point registers, but widened to
128 bits from 64. What I had tried to do was to keep the layout of the
bottom part of the structure unchanged for compatibility by storing only
the extra half of registers 1-32 in a separate area. In particular, I
wanted to keep the FP registers readable in a consecutive way.
-Nathan
___
svn-src-all@freebsd.org mailing list
http://lists.freebsd.org/mailman/listin
svn commit: r279957 - head/usr.sbin/bhyve
Author: mav
Date: Fri Mar 13 16:43:52 2015
New Revision: 279957
URL: https://svnweb.freebsd.org/changeset/base/279957
Log:
Add DSM TRIM command support for virtual AHCI disks.
It works only for virtual disks backed by ZVOLs and raw devices supporting
BIO_DELETE. Virtual disks backed by files won't report this capability.
MFC after:2 weeks
Relnotes: yes
Modified:
head/usr.sbin/bhyve/block_if.c
head/usr.sbin/bhyve/block_if.h
head/usr.sbin/bhyve/pci_ahci.c
Modified: head/usr.sbin/bhyve/block_if.c
==
--- head/usr.sbin/bhyve/block_if.c Fri Mar 13 13:46:50 2015
(r279956)
+++ head/usr.sbin/bhyve/block_if.c Fri Mar 13 16:43:52 2015
(r279957)
@@ -59,7 +59,8 @@ __FBSDID("$FreeBSD$");
enum blockop {
BOP_READ,
BOP_WRITE,
- BOP_FLUSH
+ BOP_FLUSH,
+ BOP_DELETE
};
enum blockstat {
@@ -81,6 +82,7 @@ struct blockif_ctxt {
int bc_magic;
int bc_fd;
int bc_ischr;
+ int bc_candelete;
int bc_rdonly;
off_t bc_size;
int bc_sectsz;
@@ -172,6 +174,7 @@ static void
blockif_proc(struct blockif_ctxt *bc, struct blockif_elem *be)
{
struct blockif_req *br;
+ off_t arg[2];
int err;
br = be->be_req;
@@ -197,6 +200,17 @@ blockif_proc(struct blockif_ctxt *bc, st
} else if (fsync(bc->bc_fd))
err = errno;
break;
+ case BOP_DELETE:
+ if (!bc->bc_candelete)
+ err = EOPNOTSUPP;
+ else if (bc->bc_ischr) {
+ arg[0] = br->br_offset;
+ arg[1] = br->br_iov[0].iov_len;
+ if (ioctl(bc->bc_fd, DIOCGDELETE, arg))
+ err = errno;
+ } else
+ err = EOPNOTSUPP;
+ break;
default:
err = EINVAL;
break;
@@ -276,9 +290,10 @@ blockif_open(const char *optstr, const c
char *nopt, *xopts;
struct blockif_ctxt *bc;
struct stat sbuf;
+ struct diocgattr_arg arg;
off_t size, psectsz, psectoff;
int extra, fd, i, sectsz;
- int nocache, sync, ro;
+ int nocache, sync, ro, candelete;
pthread_once(&blockif_once, blockif_init);
@@ -332,6 +347,7 @@ blockif_open(const char *optstr, const c
size = sbuf.st_size;
sectsz = DEV_BSIZE;
psectsz = psectoff = 0;
+ candelete = 0;
if (S_ISCHR(sbuf.st_mode)) {
if (ioctl(fd, DIOCGMEDIASIZE, &size) < 0 ||
ioctl(fd, DIOCGSECTORSIZE, §sz)) {
@@ -343,6 +359,10 @@ blockif_open(const char *optstr, const c
assert(sectsz != 0);
if (ioctl(fd, DIOCGSTRIPESIZE, &psectsz) == 0 && psectsz > 0)
ioctl(fd, DIOCGSTRIPEOFFSET, &psectoff);
+ strlcpy(arg.name, "GEOM::candelete", sizeof(arg.name));
+ arg.len = sizeof(arg.value.i);
+ if (ioctl(fd, DIOCGATTR, &arg) == 0)
+ candelete = arg.value.i;
} else
psectsz = sbuf.st_blksize;
@@ -355,6 +375,7 @@ blockif_open(const char *optstr, const c
bc->bc_magic = BLOCKIF_SIG;
bc->bc_fd = fd;
bc->bc_ischr = S_ISCHR(sbuf.st_mode);
+ bc->bc_candelete = candelete;
bc->bc_rdonly = ro;
bc->bc_size = size;
bc->bc_sectsz = sectsz;
@@ -434,6 +455,14 @@ blockif_flush(struct blockif_ctxt *bc, s
}
int
+blockif_delete(struct blockif_ctxt *bc, struct blockif_req *breq)
+{
+
+ assert(bc->bc_magic == BLOCKIF_SIG);
+ return (blockif_request(bc, breq, BOP_DELETE));
+}
+
+int
blockif_cancel(struct blockif_ctxt *bc, struct blockif_req *breq)
{
struct blockif_elem *be;
@@ -634,3 +663,11 @@ blockif_is_ro(struct blockif_ctxt *bc)
assert(bc->bc_magic == BLOCKIF_SIG);
return (bc->bc_rdonly);
}
+
+int
+blockif_candelete(struct blockif_ctxt *bc)
+{
+
+ assert(bc->bc_magic == BLOCKIF_SIG);
+ return (bc->bc_candelete);
+}
Modified: head/usr.sbin/bhyve/block_if.h
==
--- head/usr.sbin/bhyve/block_if.h Fri Mar 13 13:46:50 2015
(r279956)
+++ head/usr.sbin/bhyve/block_if.h Fri Mar 13 16:43:52 2015
(r279957)
@@ -58,9 +58,11 @@ int blockif_sectsz(struct blockif_ctxt *
void blockif_psectsz(struct blockif_ctxt *bc, int *size, int *off);
intblockif_queuesz(struct blockif_ctxt *bc);
intblockif_is_ro(struct blockif_ctxt *bc);
+intblockif_candelete(struct blockif_ctxt *bc);
intblockif_read(struct blockif_ctxt *bc, struct blockif_req *breq);
intblockif_write(st
Re: svn commit: r279932 - head/sys/vm
On Fri, 2015-03-13 at 06:24 -0400, John Baldwin wrote: > On Thursday, March 12, 2015 05:24:51 PM Ian Lepore wrote: > > On Thu, 2015-03-12 at 17:02 -0400, Ryan Stone wrote: > > > On Thu, Mar 12, 2015 at 2:06 PM, Ian Lepore wrote: > > > > > > > Nullterminate strings returned via sysctl. > > > > > > > > PR: 195668 > > > > > > > > > > To quote the manpage: > > > > > > > The *sbuf* family of functions allows one to safely > > > > allocate, construct and release bounded null-terminated > > > > strings in kernel space. > > > > > > IMO the sbuf API is broken if we have to explicitly null-terminate the > > > string ourselves. > > > > If we want the nullterm to be counted in the length of data in the > > buffer (and thus get transmitted back across the syscall boundary) we > > need to put an explicit counted nullterm byte into the buffer. > > > > I had started down the path of making that (counting the nullterm byte > > as part of the data in the buffer) a feature of sbuf that you could set > > with a flag, but then whoever added sbuf_new_for_sysctl() didn't > > propagate the flags field through the new function and I decided to not > > go off into the weeds making a new flavor of that takes flags. > > One suggestion would be to consider using '\0' for a nul character instead of > a bare 0. To me that communicates the intention more clearly to the reader. > (One of the things I did not like about C++ < C++11 was the use of 0 for > NULL. I much prefer nullptr and NULL in C over bare 0's for pointers for > similar clarity reasons.) > I have waffled back and forth between preferring 0 or '\0' for 30 years, I just seem to go through phases with nullterm expression. In general I'm glad I got called away to an onsite meeting yesterday and didn't get far with these changes, because the more I think about it, the less satisfied I am with this expedient fix. The other fix I started on, where a new SBUF_COUNTNUL flag can be set to inform the sbuf_finish() code that you want the terminating nul counted in the data length just feels like a better fit for the overall "automaticness" of how the sbuf stuff works. -- Ian ___ svn-src-all@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
Re: svn commit: r279937 - in head/sys/powerpc: include powerpc
On 03/13/15 01:47, Konstantin Belousov wrote:
On Thu, Mar 12, 2015 at 02:46:53PM -0700, Nathan Whitehorn wrote:
On 03/12/15 14:35, Konstantin Belousov wrote:
On Thu, Mar 12, 2015 at 02:29:43PM -0700, Nathan Whitehorn wrote:
On 03/12/15 14:22, Konstantin Belousov wrote:
On Thu, Mar 12, 2015 at 09:15:39PM +, Nathan Whitehorn wrote:
Author: nwhitehorn
Date: Thu Mar 12 21:15:38 2015
New Revision: 279937
URL: https://svnweb.freebsd.org/changeset/base/279937
Log:
Provide VSX context in ucontext(3) API.
Modified:
head/sys/powerpc/include/ucontext.h
head/sys/powerpc/powerpc/exec_machdep.c
Modified: head/sys/powerpc/include/ucontext.h
==
--- head/sys/powerpc/include/ucontext.h Thu Mar 12 20:14:48 2015
(r279936)
+++ head/sys/powerpc/include/ucontext.h Thu Mar 12 21:15:38 2015
(r279937)
@@ -46,6 +46,7 @@ typedef struct __mcontext {
uint32_tmc_av[2];
register_t mc_frame[42];
uint64_tmc_fpreg[33];
+ uint64_tmc_vsxfpreg[32];/* low-order half of VSR0-31 */
} mcontext_t __aligned(16);
#if defined(_KERNEL) && defined(__powerpc64__)
@@ -60,6 +61,7 @@ typedef struct __mcontext32 {
uint32_tmc_av[2];
uint32_tmc_frame[42];
uint64_tmc_fpreg[33];
+ uint64_tmc_vsxfpreg[32];/* low-order half of VSR0-31 */
} mcontext32_t __aligned(16);
#endif
It looks as if you broken the ABI compatibility by the change. Am I wrong ?
That is correct. It's a tier-2 platform and -CURRENT, so I'm not sure
it's worth the compatibility shims. I'm happy to add them if you think
otherwise.
You are main maintainer of PowerPC port, IMO, so it is your decision.
Note that 'this is current' argument is not applicable, since the change
also breaks stable/* binaries.
I do understand the argument of PowerPC being tier 2 architecture, but this
makes me sad. Anyway, it is yours. For x86, I have to introduce
getcontextx(3) mechanism.
This is a good point. I'll try to fix it. Is my understanding of how
this works correct?
1. Provide a sysarch() for the extended FPU state.
2. Implement getcontextx() in the C library to fill extra properties if
required.
3. Store state for signal trampoline in variable-sized stack area
4. Implement __getcontextx_size() and __fillcontextx2() for use
in the deferred signal delivery while libthr is in critical section,
see lib/libthr/thread/thr_sig.c:check_deferred_signal().
Implementation of (2) seems to rely on having spare members in ucontext,
which PowerPC unfortunately does not have. Is there a way around that?
Indeed, this is very unfortunate.
One more thing I'd appreciate advice on: we have a much worse ABI issue
with setjmp(). On 32-bit systems, _JBLEN is too small to hold all the
nonvolatile vector registers. Do you have any ideas how to handle that?
-Nathan
___
svn-src-all@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/svn-src-all
To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
Re: svn commit: r279932 - head/sys/vm
On Friday, March 13, 2015 10:14:27 AM Ian Lepore wrote: > On Fri, 2015-03-13 at 06:24 -0400, John Baldwin wrote: > > On Thursday, March 12, 2015 05:24:51 PM Ian Lepore wrote: > > > On Thu, 2015-03-12 at 17:02 -0400, Ryan Stone wrote: > > > > On Thu, Mar 12, 2015 at 2:06 PM, Ian Lepore wrote: > > > > > > > > > Nullterminate strings returned via sysctl. > > > > > > > > > > PR: 195668 > > > > > > > > > > > > > To quote the manpage: > > > > > > > > > The *sbuf* family of functions allows one to safely > > > > > allocate, construct and release bounded null-terminated > > > > > strings in kernel space. > > > > > > > > IMO the sbuf API is broken if we have to explicitly null-terminate the > > > > string ourselves. > > > > > > If we want the nullterm to be counted in the length of data in the > > > buffer (and thus get transmitted back across the syscall boundary) we > > > need to put an explicit counted nullterm byte into the buffer. > > > > > > I had started down the path of making that (counting the nullterm byte > > > as part of the data in the buffer) a feature of sbuf that you could set > > > with a flag, but then whoever added sbuf_new_for_sysctl() didn't > > > propagate the flags field through the new function and I decided to not > > > go off into the weeds making a new flavor of that takes flags. > > > > One suggestion would be to consider using '\0' for a nul character instead > > of > > a bare 0. To me that communicates the intention more clearly to the reader. > > (One of the things I did not like about C++ < C++11 was the use of 0 for > > NULL. I much prefer nullptr and NULL in C over bare 0's for pointers for > > similar clarity reasons.) > > > > I have waffled back and forth between preferring 0 or '\0' for 30 years, > I just seem to go through phases with nullterm expression. > > In general I'm glad I got called away to an onsite meeting yesterday and > didn't get far with these changes, because the more I think about it, > the less satisfied I am with this expedient fix. The other fix I > started on, where a new SBUF_COUNTNUL flag can be set to inform the > sbuf_finish() code that you want the terminating nul counted in the data > length just feels like a better fit for the overall "automaticness" of > how the sbuf stuff works. Hmm, I actually think that it's a bug that the terminating nul isn't included when draining. If we fixed that then I think that fixes most of these? The places that explicitly use 'sysctl_handle_string()' with an sbuf should probably just be using sbuf_len(sb) + 1' explicitly. (Another option would be to have a sysctl_handle_sbuf() that was a wrapper around sysctl_handle_string() that included the + 1 to hide that detail if there is more than one.) -- John Baldwin ___ svn-src-all@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
Re: svn commit: r279937 - in head/sys/powerpc: include powerpc
On Fri, Mar 13, 2015 at 10:06:05AM -0700, Nathan Whitehorn wrote: > One more thing I'd appreciate advice on: we have a much worse ABI issue > with setjmp(). On 32-bit systems, _JBLEN is too small to hold all the > nonvolatile vector registers. Do you have any ideas how to handle that? There is no solution there. On x86, both 32 and 64 bit, setjmp only saves fpu control word (and mxcsr on amd64, where SSE2 is guaranteed to be presented). So only limited bits of the whole arch state is saved and restored. I suggest to just leave setjmp as it is. In fact, FreeBSD cannot work on 486SX due to this. Shell uses setjmp, and despite kernel carefully handling FPU-less CPU, shell fails from the beginning. Please note that ucontext_t API was designed by System V group to handle this deficiency in the setjmp/longjmp API, to cover the whole machine context and provide almost-portable interface for non-local control flow :/. As usual, SysV produced something which failed. In particular, the size and layout of ucontext_t is user-visible and cannot be changed. ___ svn-src-all@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r279958 - in stable: 10/share/mk 9/share/mk
Author: jhb
Date: Fri Mar 13 17:45:34 2015
New Revision: 279958
URL: https://svnweb.freebsd.org/changeset/base/279958
Log:
MFC 278682:
Make the extra dependencies in DPADD be dependencies of PROG_FULL and
SHLIB_NAME_FULL so that the full binary is relinked when a dependency
changes. Right now the existing full binary is left as-is and only
the objcopy to remove debug symbols is run.
Modified:
stable/9/share/mk/bsd.lib.mk
stable/9/share/mk/bsd.prog.mk
Directory Properties:
stable/9/share/mk/ (props changed)
Changes in other areas also in this revision:
Modified:
stable/10/share/mk/bsd.lib.mk
stable/10/share/mk/bsd.prog.mk
Directory Properties:
stable/10/ (props changed)
Modified: stable/9/share/mk/bsd.lib.mk
==
--- stable/9/share/mk/bsd.lib.mkFri Mar 13 16:43:52 2015
(r279957)
+++ stable/9/share/mk/bsd.lib.mkFri Mar 13 17:45:34 2015
(r279958)
@@ -275,7 +275,7 @@ _EXTRADEPEND:
mv $$TMP ${DEPENDFILE}
.if !defined(NO_EXTRADEPEND) && defined(SHLIB_NAME)
.if defined(DPADD) && !empty(DPADD)
- echo ${SHLIB_NAME}: ${DPADD} >> ${DEPENDFILE}
+ echo ${SHLIB_NAME_FULL}: ${DPADD} >> ${DEPENDFILE}
.endif
.endif
Modified: stable/9/share/mk/bsd.prog.mk
==
--- stable/9/share/mk/bsd.prog.mk Fri Mar 13 16:43:52 2015
(r279957)
+++ stable/9/share/mk/bsd.prog.mk Fri Mar 13 17:45:34 2015
(r279958)
@@ -168,15 +168,15 @@ CLEANFILES+= ${OBJS}
_EXTRADEPEND:
.if defined(LDFLAGS) && !empty(LDFLAGS:M-nostdlib)
.if defined(DPADD) && !empty(DPADD)
- echo ${PROG}: ${DPADD} >> ${DEPENDFILE}
+ echo ${PROG_FULL}: ${DPADD} >> ${DEPENDFILE}
.endif
.else
- echo ${PROG}: ${LIBC} ${DPADD} >> ${DEPENDFILE}
+ echo ${PROG_FULL}: ${LIBC} ${DPADD} >> ${DEPENDFILE}
.if defined(PROG_CXX) && !defined(EARLY_BUILD)
.if !empty(CXXFLAGS:M-stdlib=libc++)
- echo ${PROG}: ${LIBCPLUSPLUS} >> ${DEPENDFILE}
+ echo ${PROG_FULL}: ${LIBCPLUSPLUS} >> ${DEPENDFILE}
.else
- echo ${PROG}: ${LIBSTDCPLUSPLUS} >> ${DEPENDFILE}
+ echo ${PROG_FULL}: ${LIBSTDCPLUSPLUS} >> ${DEPENDFILE}
.endif
.endif
.endif
___
svn-src-all@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/svn-src-all
To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
Re: svn commit: r279937 - in head/sys/powerpc: include powerpc
On 03/13/15 10:23, Konstantin Belousov wrote: On Fri, Mar 13, 2015 at 10:06:05AM -0700, Nathan Whitehorn wrote: One more thing I'd appreciate advice on: we have a much worse ABI issue with setjmp(). On 32-bit systems, _JBLEN is too small to hold all the nonvolatile vector registers. Do you have any ideas how to handle that? There is no solution there. On x86, both 32 and 64 bit, setjmp only saves fpu control word (and mxcsr on amd64, where SSE2 is guaranteed to be presented). So only limited bits of the whole arch state is saved and restored. I suggest to just leave setjmp as it is. In fact, FreeBSD cannot work on 486SX due to this. Shell uses setjmp, and despite kernel carefully handling FPU-less CPU, shell fails from the beginning. Please note that ucontext_t API was designed by System V group to handle this deficiency in the setjmp/longjmp API, to cover the whole machine context and provide almost-portable interface for non-local control flow :/. As usual, SysV produced something which failed. In particular, the size and layout of ucontext_t is user-visible and cannot be changed. We'll need to hack the compiler in this case, since it assumes setjmp() saves and restores the vector registers. I'm really not sure which option is worse. -Nathan ___ svn-src-all@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r279958 - in stable: 10/share/mk 9/share/mk
Author: jhb
Date: Fri Mar 13 17:45:34 2015
New Revision: 279958
URL: https://svnweb.freebsd.org/changeset/base/279958
Log:
MFC 278682:
Make the extra dependencies in DPADD be dependencies of PROG_FULL and
SHLIB_NAME_FULL so that the full binary is relinked when a dependency
changes. Right now the existing full binary is left as-is and only
the objcopy to remove debug symbols is run.
Modified:
stable/10/share/mk/bsd.lib.mk
stable/10/share/mk/bsd.prog.mk
Directory Properties:
stable/10/ (props changed)
Changes in other areas also in this revision:
Modified:
stable/9/share/mk/bsd.lib.mk
stable/9/share/mk/bsd.prog.mk
Directory Properties:
stable/9/share/mk/ (props changed)
Modified: stable/10/share/mk/bsd.lib.mk
==
--- stable/10/share/mk/bsd.lib.mk Fri Mar 13 16:43:52 2015
(r279957)
+++ stable/10/share/mk/bsd.lib.mk Fri Mar 13 17:45:34 2015
(r279958)
@@ -276,7 +276,7 @@ _EXTRADEPEND:
mv $$TMP ${DEPENDFILE}
.if !defined(NO_EXTRADEPEND) && defined(SHLIB_NAME)
.if defined(DPADD) && !empty(DPADD)
- echo ${SHLIB_NAME}: ${DPADD} >> ${DEPENDFILE}
+ echo ${SHLIB_NAME_FULL}: ${DPADD} >> ${DEPENDFILE}
.endif
.endif
Modified: stable/10/share/mk/bsd.prog.mk
==
--- stable/10/share/mk/bsd.prog.mk Fri Mar 13 16:43:52 2015
(r279957)
+++ stable/10/share/mk/bsd.prog.mk Fri Mar 13 17:45:34 2015
(r279958)
@@ -167,15 +167,15 @@ CLEANFILES+= ${OBJS}
_EXTRADEPEND:
.if defined(LDFLAGS) && !empty(LDFLAGS:M-nostdlib)
.if defined(DPADD) && !empty(DPADD)
- echo ${PROG}: ${DPADD} >> ${DEPENDFILE}
+ echo ${PROG_FULL}: ${DPADD} >> ${DEPENDFILE}
.endif
.else
- echo ${PROG}: ${LIBC} ${DPADD} >> ${DEPENDFILE}
+ echo ${PROG_FULL}: ${LIBC} ${DPADD} >> ${DEPENDFILE}
.if defined(PROG_CXX)
.if ${MK_CLANG_IS_CC} != "no" && empty(CXXFLAGS:M-stdlib=libstdc++)
- echo ${PROG}: ${LIBCPLUSPLUS} >> ${DEPENDFILE}
+ echo ${PROG_FULL}: ${LIBCPLUSPLUS} >> ${DEPENDFILE}
.else
- echo ${PROG}: ${LIBSTDCPLUSPLUS} >> ${DEPENDFILE}
+ echo ${PROG_FULL}: ${LIBSTDCPLUSPLUS} >> ${DEPENDFILE}
.endif
.endif
.endif
___
svn-src-all@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/svn-src-all
To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
Re: svn commit: r279932 - head/sys/vm
On Fri, 2015-03-13 at 13:19 -0400, John Baldwin wrote: > On Friday, March 13, 2015 10:14:27 AM Ian Lepore wrote: > > On Fri, 2015-03-13 at 06:24 -0400, John Baldwin wrote: > > > On Thursday, March 12, 2015 05:24:51 PM Ian Lepore wrote: [...] > > > > In general I'm glad I got called away to an onsite meeting yesterday and > > didn't get far with these changes, because the more I think about it, > > the less satisfied I am with this expedient fix. The other fix I > > started on, where a new SBUF_COUNTNUL flag can be set to inform the > > sbuf_finish() code that you want the terminating nul counted in the data > > length just feels like a better fit for the overall "automaticness" of > > how the sbuf stuff works. > > Hmm, I actually think that it's a bug that the terminating nul isn't included > when draining. If we fixed that then I think that fixes most of these? > The places that explicitly use 'sysctl_handle_string()' with an sbuf > should probably just be using sbuf_len(sb) + 1' explicitly. (Another > option would be to have a sysctl_handle_sbuf() that was a wrapper around > sysctl_handle_string() that included the + 1 to hide that detail if there is > more than one.) > Some of the uses of sbuf for sysctl use sbuf_bcat() for dealing with binary structs, so we can't just assume that a nullterm should be added and included in the buffer length -- there needs to be some mechanism to say explicitly "this is an sbuf for a sysctl string" (and more generally "this is an sbuf for a string where I want the nul byte counted as part of the data" because that could be useful in non-sysctl contexts too, especially in userland). -- Ian ___ svn-src-all@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
Re: svn commit: r279937 - in head/sys/powerpc: include powerpc
On Fri, Mar 13, 2015 at 10:45:46AM -0700, Nathan Whitehorn wrote: > We'll need to hack the compiler in this case, since it assumes setjmp() > saves and restores the vector registers. I'm really not sure which > option is worse. Changing the compiler is arguably much worse than breaking ABI of the tier 2 platform, indeed. We must maintain the situation where the stock build of the compilers work out of box. Still, how the compiler' assumptions are laid out ? It could be argued that compilers on x86 also assume that FPU register file is restored by longjmp. %st* and %xmm* are defined as not preserved across function calls, but I suspect that practical rule for setjmp() is that floating vars better not be used in the target frame. Hm, indeed f14-f31 and v20-v31 are marked as non-volatile for 64bit ABI. ___ svn-src-all@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r279959 - head/usr.sbin/bhyve
Author: mav
Date: Fri Mar 13 18:04:07 2015
New Revision: 279959
URL: https://svnweb.freebsd.org/changeset/base/279959
Log:
Use ahci_write_fis_d2h() for commands completion.
MFC after:2 weeks
Modified:
head/usr.sbin/bhyve/pci_ahci.c
Modified: head/usr.sbin/bhyve/pci_ahci.c
==
--- head/usr.sbin/bhyve/pci_ahci.c Fri Mar 13 17:45:34 2015
(r279958)
+++ head/usr.sbin/bhyve/pci_ahci.c Fri Mar 13 18:04:07 2015
(r279959)
@@ -773,8 +773,8 @@ handle_identify(struct ahci_port *p, int
hdr = (struct ahci_cmd_hdr *)(p->cmd_lst + slot * AHCI_CL_SIZE);
if (p->atapi || hdr->prdtl == 0) {
- p->tfd = (ATA_E_ABORT << 8) | ATA_S_READY | ATA_S_ERROR;
- p->is |= AHCI_P_IX_TFE;
+ ahci_write_fis_d2h(p, slot, cfis,
+ (ATA_E_ABORT << 8) | ATA_S_READY | ATA_S_ERROR);
} else {
uint16_t buf[256];
uint64_t sectors;
@@ -851,19 +851,16 @@ handle_identify(struct ahci_port *p, int
}
ahci_write_fis_piosetup(p);
write_prdt(p, slot, cfis, (void *)buf, sizeof(buf));
- p->tfd = ATA_S_DSC | ATA_S_READY;
- p->is |= AHCI_P_IX_DP;
- p->ci &= ~(1 << slot);
+ ahci_write_fis_d2h(p, slot, cfis, ATA_S_DSC | ATA_S_READY);
}
- ahci_generate_intr(p->pr_sc);
}
static void
handle_atapi_identify(struct ahci_port *p, int slot, uint8_t *cfis)
{
if (!p->atapi) {
- p->tfd = (ATA_E_ABORT << 8) | ATA_S_READY | ATA_S_ERROR;
- p->is |= AHCI_P_IX_TFE;
+ ahci_write_fis_d2h(p, slot, cfis,
+ (ATA_E_ABORT << 8) | ATA_S_READY | ATA_S_ERROR);
} else {
uint16_t buf[256];
@@ -894,11 +891,8 @@ handle_atapi_identify(struct ahci_port *
buf[88] = (1 << 14 | 0x7f);
ahci_write_fis_piosetup(p);
write_prdt(p, slot, cfis, (void *)buf, sizeof(buf));
- p->tfd = ATA_S_DSC | ATA_S_READY;
- p->is |= AHCI_P_IX_DHR;
- p->ci &= ~(1 << slot);
+ ahci_write_fis_d2h(p, slot, cfis, ATA_S_DSC | ATA_S_READY);
}
- ahci_generate_intr(p->pr_sc);
}
static void
@@ -1479,9 +1473,7 @@ ahci_handle_cmd(struct ahci_port *p, int
p->mult_sectors = cfis[12];
p->tfd = ATA_S_DSC | ATA_S_READY;
}
- p->is |= AHCI_P_IX_DP;
- p->ci &= ~(1 << slot);
- ahci_generate_intr(p->pr_sc);
+ ahci_write_fis_d2h(p, slot, cfis, p->tfd);
break;
case ATA_READ_DMA:
case ATA_WRITE_DMA:
@@ -1517,17 +1509,15 @@ ahci_handle_cmd(struct ahci_port *p, int
break;
case ATA_PACKET_CMD:
if (!p->atapi) {
- p->tfd = (ATA_E_ABORT << 8) | ATA_S_READY | ATA_S_ERROR;
- p->is |= AHCI_P_IX_TFE;
- ahci_generate_intr(p->pr_sc);
+ ahci_write_fis_d2h(p, slot, cfis,
+ (ATA_E_ABORT << 8) | ATA_S_READY | ATA_S_ERROR);
} else
handle_packet_cmd(p, slot, cfis);
break;
default:
WPRINTF("Unsupported cmd:%02x\n", cfis[2]);
- p->tfd = (ATA_E_ABORT << 8) | ATA_S_READY | ATA_S_ERROR;
- p->is |= AHCI_P_IX_TFE;
- ahci_generate_intr(p->pr_sc);
+ ahci_write_fis_d2h(p, slot, cfis,
+ (ATA_E_ABORT << 8) | ATA_S_READY | ATA_S_ERROR);
break;
}
}
___
svn-src-all@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/svn-src-all
To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
Re: svn commit: r279937 - in head/sys/powerpc: include powerpc
On 03/13/15 10:59, Konstantin Belousov wrote: On Fri, Mar 13, 2015 at 10:45:46AM -0700, Nathan Whitehorn wrote: We'll need to hack the compiler in this case, since it assumes setjmp() saves and restores the vector registers. I'm really not sure which option is worse. Changing the compiler is arguably much worse than breaking ABI of the tier 2 platform, indeed. We must maintain the situation where the stock build of the compilers work out of box. Still, how the compiler' assumptions are laid out ? It could be argued that compilers on x86 also assume that FPU register file is restored by longjmp. %st* and %xmm* are defined as not preserved across function calls, but I suspect that practical rule for setjmp() is that floating vars better not be used in the target frame. Hm, indeed f14-f31 and v20-v31 are marked as non-volatile for 64bit ABI. The compiler assumes that f14-f31 and v20-v31 are preserved by setjmp() (and, e.g. glibc's setjmp() does in fact preserve them). It also tries not to use those registers too much, so the practical impact is limited, but it should actually be fixed. -Nathan ___ svn-src-all@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r279960 - head/usr.sbin/bhyve
Author: mav
Date: Fri Mar 13 18:35:38 2015
New Revision: 279960
URL: https://svnweb.freebsd.org/changeset/base/279960
Log:
Add support for PIO variants of READ/WRITE commands for AHCI disks.
AHCI API hides all PIO specifics, so this functionality is almost free.
MFC after:2 weeks
Modified:
head/usr.sbin/bhyve/pci_ahci.c
Modified: head/usr.sbin/bhyve/pci_ahci.c
==
--- head/usr.sbin/bhyve/pci_ahci.c Fri Mar 13 18:04:07 2015
(r279959)
+++ head/usr.sbin/bhyve/pci_ahci.c Fri Mar 13 18:35:38 2015
(r279960)
@@ -520,12 +520,14 @@ ahci_handle_dma(struct ahci_port *p, int
readop = 1;
prdt += seek;
- if (cfis[2] == ATA_WRITE_DMA || cfis[2] == ATA_WRITE_DMA48 ||
- cfis[2] == ATA_WRITE_FPDMA_QUEUED)
+ if (cfis[2] == ATA_WRITE || cfis[2] == ATA_WRITE48 ||
+ cfis[2] == ATA_WRITE_MUL || cfis[2] == ATA_WRITE_MUL48 ||
+ cfis[2] == ATA_WRITE_DMA || cfis[2] == ATA_WRITE_DMA48 ||
+ cfis[2] == ATA_WRITE_FPDMA_QUEUED)
readop = 0;
if (cfis[2] == ATA_WRITE_FPDMA_QUEUED ||
- cfis[2] == ATA_READ_FPDMA_QUEUED) {
+ cfis[2] == ATA_READ_FPDMA_QUEUED) {
lba = ((uint64_t)cfis[10] << 40) |
((uint64_t)cfis[9] << 32) |
((uint64_t)cfis[8] << 24) |
@@ -536,7 +538,9 @@ ahci_handle_dma(struct ahci_port *p, int
if (!len)
len = 65536;
ncq = 1;
- } else if (cfis[2] == ATA_READ_DMA48 || cfis[2] == ATA_WRITE_DMA48) {
+ } else if (cfis[2] == ATA_READ48 || cfis[2] == ATA_WRITE48 ||
+ cfis[2] == ATA_READ_MUL48 || cfis[2] == ATA_WRITE_MUL48 ||
+ cfis[2] == ATA_READ_DMA48 || cfis[2] == ATA_WRITE_DMA48) {
lba = ((uint64_t)cfis[10] << 40) |
((uint64_t)cfis[9] << 32) |
((uint64_t)cfis[8] << 24) |
@@ -1475,6 +1479,14 @@ ahci_handle_cmd(struct ahci_port *p, int
}
ahci_write_fis_d2h(p, slot, cfis, p->tfd);
break;
+ case ATA_READ:
+ case ATA_WRITE:
+ case ATA_READ48:
+ case ATA_WRITE48:
+ case ATA_READ_MUL:
+ case ATA_WRITE_MUL:
+ case ATA_READ_MUL48:
+ case ATA_WRITE_MUL48:
case ATA_READ_DMA:
case ATA_WRITE_DMA:
case ATA_READ_DMA48:
___
svn-src-all@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/svn-src-all
To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r279961 - stable/10/sys/dev/drm2/i915
Author: jhb
Date: Fri Mar 13 18:38:02 2015
New Revision: 279961
URL: https://svnweb.freebsd.org/changeset/base/279961
Log:
MFC 270516:
i915 driver - enable opregion handle; program CADL.
add opregion handling for drm2 - which exposes some ACPI video configuration
pieces that some Lenovo laptop models use to flesh out which video device
to speak to. This enables the brightness control in ACPI to work these
models.
The CADL bits are also important - it's used to figure out which ACPI
events to hook the brightness buttons into. It doesn't yet seem to work
for me, but it does for the OP.
PR: 190186, 198551
Submitted by: Henry Hu
Modified:
stable/10/sys/dev/drm2/i915/i915_drv.h
stable/10/sys/dev/drm2/i915/i915_irq.c
stable/10/sys/dev/drm2/i915/intel_opregion.c
Directory Properties:
stable/10/ (props changed)
Modified: stable/10/sys/dev/drm2/i915/i915_drv.h
==
--- stable/10/sys/dev/drm2/i915/i915_drv.h Fri Mar 13 18:35:38 2015
(r279960)
+++ stable/10/sys/dev/drm2/i915/i915_drv.h Fri Mar 13 18:38:02 2015
(r279961)
@@ -1277,10 +1277,11 @@ extern void intel_iic_reset(struct drm_d
/* intel_opregion.c */
int intel_opregion_setup(struct drm_device *dev);
-extern int intel_opregion_init(struct drm_device *dev);
+extern void intel_opregion_init(struct drm_device *dev);
extern void intel_opregion_fini(struct drm_device *dev);
-extern void opregion_asle_intr(struct drm_device *dev);
-extern void opregion_enable_asle(struct drm_device *dev);
+extern void intel_opregion_asle_intr(struct drm_device *dev);
+extern void intel_opregion_gse_intr(struct drm_device *dev);
+extern void intel_opregion_enable_asle(struct drm_device *dev);
/* i915_gem_gtt.c */
int i915_gem_init_aliasing_ppgtt(struct drm_device *dev);
Modified: stable/10/sys/dev/drm2/i915/i915_irq.c
==
--- stable/10/sys/dev/drm2/i915/i915_irq.c Fri Mar 13 18:35:38 2015
(r279960)
+++ stable/10/sys/dev/drm2/i915/i915_irq.c Fri Mar 13 18:38:02 2015
(r279961)
@@ -537,11 +537,7 @@ ivybridge_irq_handler(void *arg)
notify_ring(dev, &dev_priv->rings[BCS]);
if (de_iir & DE_GSE_IVB) {
-#if 1
- KIB_NOTYET();
-#else
intel_opregion_gse_intr(dev);
-#endif
}
if (de_iir & DE_PLANEA_FLIP_DONE_IVB) {
@@ -649,11 +645,7 @@ ironlake_irq_handler(void *arg)
notify_ring(dev, &dev_priv->rings[BCS]);
if (de_iir & DE_GSE) {
-#if 1
- KIB_NOTYET();
-#else
intel_opregion_gse_intr(dev);
-#endif
}
if (de_iir & DE_PLANEA_FLIP_DONE) {
@@ -1055,11 +1047,7 @@ i915_driver_irq_handler(void *arg)
if (blc_event || (iir & I915_ASLE_INTERRUPT)) {
-#if 1
- KIB_NOTYET();
-#else
intel_opregion_asle_intr(dev);
-#endif
}
/* With MSI, interrupts are only generated when iir
@@ -1781,11 +1769,7 @@ i915_driver_irq_postinstall(struct drm_d
I915_WRITE(PORT_HOTPLUG_EN, hotplug_en);
}
-#if 1
- KIB_NOTYET();
-#else
intel_opregion_enable_asle(dev);
-#endif
return 0;
}
Modified: stable/10/sys/dev/drm2/i915/intel_opregion.c
==
--- stable/10/sys/dev/drm2/i915/intel_opregion.cFri Mar 13 18:35:38
2015(r279960)
+++ stable/10/sys/dev/drm2/i915/intel_opregion.cFri Mar 13 18:38:02
2015(r279961)
@@ -32,6 +32,9 @@ __FBSDID("$FreeBSD$");
#include
#include
#include
+#include
+#include
+#include
#define PCI_ASLE 0xe4
#define PCI_ASLS 0xfc
@@ -144,7 +147,7 @@ struct opregion_asle {
#define ACPI_DIGITAL_OUTPUT (3<<8)
#define ACPI_LVDS_OUTPUT (4<<8)
-#ifdef CONFIG_ACPI
+#if 1
static u32 asle_set_backlight(struct drm_device *dev, u32 bclp)
{
struct drm_i915_private *dev_priv = dev->dev_private;
@@ -289,6 +292,7 @@ void intel_opregion_enable_asle(struct d
static struct intel_opregion *system_opregion;
+#if 0
static int intel_opregion_video_event(struct notifier_block *nb,
unsigned long val, void *data)
{
@@ -319,6 +323,7 @@ static int intel_opregion_video_event(st
static struct notifier_block intel_opregion_notifier = {
.notifier_call = intel_opregion_video_event,
};
+#endif
/*
* Initialise the DIDL field in opregion. This passes a list of devices to
@@ -326,37 +331,72 @@ static struct notifier_block intel_opreg
* (version 3)
*/
+static int acpi_is_video_device(ACPI_HANDLE devh) {
+ ACPI_HANDLE h;
+ if (ACPI_FAILURE(AcpiGetHandle(devh, "_DOD", &h)) ||
+ ACPI_FAILURE(AcpiGetHandle(devh, "_DOS", &h))) {
+ return 0;
+ }
+ return 1;
+}
+
svn commit: r279962 - head/lib/libc/posix1e
Author: pfg
Date: Fri Mar 13 18:42:43 2015
New Revision: 279962
URL: https://svnweb.freebsd.org/changeset/base/279962
Log:
libc: plug memory leaks in edge cases for the posix1e code.
CID: 1016705
CID: 1016706
CID: 1016707
Differential Revision:https://reviews.freebsd.org/D2023
Modified:
head/lib/libc/posix1e/acl_calc_mask.c
head/lib/libc/posix1e/acl_strip.c
Modified: head/lib/libc/posix1e/acl_calc_mask.c
==
--- head/lib/libc/posix1e/acl_calc_mask.c Fri Mar 13 18:38:02 2015
(r279961)
+++ head/lib/libc/posix1e/acl_calc_mask.c Fri Mar 13 18:42:43 2015
(r279962)
@@ -104,6 +104,7 @@ acl_calc_mask(acl_t *acl_p)
/* if no mask exists, check acl_cnt... */
if (acl_int_new->acl_cnt == ACL_MAX_ENTRIES) {
errno = ENOMEM;
+ acl_free(acl_new);
return (-1);
}
/* ...and add the mask entry */
Modified: head/lib/libc/posix1e/acl_strip.c
==
--- head/lib/libc/posix1e/acl_strip.c Fri Mar 13 18:38:02 2015
(r279961)
+++ head/lib/libc/posix1e/acl_strip.c Fri Mar 13 18:42:43 2015
(r279962)
@@ -82,8 +82,10 @@ _posix1e_acl_strip_np(const acl_t aclp,
have_mask_entry = 0;
acl_new = acl_init(ACL_MAX_ENTRIES);
- if (acl_new == NULL)
+ if (acl_new == NULL) {
+ acl_free(acl_old);
return (NULL);
+ }
tag = ACL_UNDEFINED_TAG;
/* only save the default user/group/other entries */
@@ -94,16 +96,16 @@ _posix1e_acl_strip_np(const acl_t aclp,
assert(_entry_brand(entry) == ACL_BRAND_POSIX);
if (acl_get_tag_type(entry, &tag) == -1)
- return (NULL);
+ goto fail;
switch(tag) {
case ACL_USER_OBJ:
case ACL_GROUP_OBJ:
case ACL_OTHER:
if (acl_get_tag_type(entry, &tag) == -1)
- return (NULL);
+ goto fail;
if (acl_get_permset(entry, &perm) == -1)
- return (NULL);
+ goto fail;
if (acl_create_entry(&acl_new, &entry_new) == -1)
return (NULL);
if (acl_set_tag_type(entry_new, tag) == -1)
@@ -120,6 +122,10 @@ _posix1e_acl_strip_np(const acl_t aclp,
default:
break;
}
+fail:
+ acl_free(acl_new);
+ acl_free(acl_old);
+ return (NULL);
}
assert(_acl_brand(acl_new) == ACL_BRAND_POSIX);
___
svn-src-all@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/svn-src-all
To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
Re: svn commit: r279937 - in head/sys/powerpc: include powerpc
On Fri, Mar 13, 2015 at 11:13:22AM -0700, Nathan Whitehorn wrote: > The compiler assumes that f14-f31 and v20-v31 are preserved by setjmp() > (and, e.g. glibc's setjmp() does in fact preserve them). It also tries > not to use those registers too much, so the practical impact is limited, > but it should actually be fixed. I think that the setjmp/longjmp case would benefit greatly from the versioning of the symbols. Versioning cannot solve all ABI issues, e.g. if you compiler used old machine/setjmp.h and then link resulting .o with newer libc, but for complete builds it will help and keep old binaries working not worse than now. ___ svn-src-all@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
Re: svn commit: r279932 - head/sys/vm
On Friday, March 13, 2015 11:57:58 AM Ian Lepore wrote: > On Fri, 2015-03-13 at 13:19 -0400, John Baldwin wrote: > > On Friday, March 13, 2015 10:14:27 AM Ian Lepore wrote: > > > On Fri, 2015-03-13 at 06:24 -0400, John Baldwin wrote: > > > > On Thursday, March 12, 2015 05:24:51 PM Ian Lepore wrote: > [...] > > > > > > In general I'm glad I got called away to an onsite meeting yesterday and > > > didn't get far with these changes, because the more I think about it, > > > the less satisfied I am with this expedient fix. The other fix I > > > started on, where a new SBUF_COUNTNUL flag can be set to inform the > > > sbuf_finish() code that you want the terminating nul counted in the data > > > length just feels like a better fit for the overall "automaticness" of > > > how the sbuf stuff works. > > > > Hmm, I actually think that it's a bug that the terminating nul isn't > > included > > when draining. If we fixed that then I think that fixes most of these? > > The places that explicitly use 'sysctl_handle_string()' with an sbuf > > should probably just be using sbuf_len(sb) + 1' explicitly. (Another > > option would be to have a sysctl_handle_sbuf() that was a wrapper around > > sysctl_handle_string() that included the + 1 to hide that detail if there is > > more than one.) > > > > Some of the uses of sbuf for sysctl use sbuf_bcat() for dealing with > binary structs, so we can't just assume that a nullterm should be added > and included in the buffer length -- there needs to be some mechanism to > say explicitly "this is an sbuf for a sysctl string" (and more generally > "this is an sbuf for a string where I want the nul byte counted as part > of the data" because that could be useful in non-sysctl contexts too, > especially in userland). Humm, that would seem to be an abuse of the API really. It is specifically designed for strings as someone else noted at the start of this thread (and as noted in the manpage). If anything I'd argue that the use cases that don't want a string should be the ones that should get a special flag in that case (or perhaps we should have a different little API to manage a buffer used for a draining sysctl where the data is a binary blob instead of a string). If you agree I'm happy to do some of the work (e.g. the different wrapper API). -- John Baldwin ___ svn-src-all@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r279964 - in stable: 8/sys/dev/ipmi 9/sys/dev/ipmi
Author: jhb
Date: Fri Mar 13 20:10:09 2015
New Revision: 279964
URL: https://svnweb.freebsd.org/changeset/base/279964
Log:
MFC 278321:
Use direct hardware access for internal requests for KCS and SMIC. In
particular, updates to the watchdog should no longer sleep.
- Add a new IPMI_IO_LOCK for low-level I/O access. Use this for
kcs_polled_request() and smic_polled_request().
- Add a new backend callback "ipmi_driver_request" to handle a driver
request. The new callback performs the request sychronously for KCS
and SMIC. SSIF still defers the work to the worker thread since the
worker thread sleeps during request processing anyway.
- Allocate driver requests on the stack rather than using malloc().
Modified:
stable/8/sys/dev/ipmi/ipmi.c
stable/8/sys/dev/ipmi/ipmi_kcs.c
stable/8/sys/dev/ipmi/ipmi_smic.c
stable/8/sys/dev/ipmi/ipmi_ssif.c
stable/8/sys/dev/ipmi/ipmivars.h
Directory Properties:
stable/8/sys/ (props changed)
stable/8/sys/dev/ (props changed)
stable/8/sys/dev/ipmi/ (props changed)
Changes in other areas also in this revision:
Modified:
stable/9/sys/dev/ipmi/ipmi.c
stable/9/sys/dev/ipmi/ipmi_kcs.c
stable/9/sys/dev/ipmi/ipmi_smic.c
stable/9/sys/dev/ipmi/ipmi_ssif.c
stable/9/sys/dev/ipmi/ipmivars.h
Directory Properties:
stable/9/sys/ (props changed)
stable/9/sys/dev/ (props changed)
Modified: stable/8/sys/dev/ipmi/ipmi.c
==
--- stable/8/sys/dev/ipmi/ipmi.cFri Mar 13 20:08:35 2015
(r279963)
+++ stable/8/sys/dev/ipmi/ipmi.cFri Mar 13 20:10:09 2015
(r279964)
@@ -49,6 +49,23 @@ __FBSDID("$FreeBSD$");
#include
#endif
+/*
+ * Driver request structures are allocated on the stack via alloca() to
+ * avoid calling malloc(), especially for the watchdog handler.
+ * To avoid too much stack growth, a previously allocated structure can
+ * be reused via IPMI_INIT_DRIVER_REQUEST(), but the caller should ensure
+ * that there is adequate reply/request space in the original allocation.
+ */
+#defineIPMI_INIT_DRIVER_REQUEST(req, addr, cmd, reqlen, replylen)
\
+ bzero((req), sizeof(struct ipmi_request)); \
+ ipmi_init_request((req), NULL, 0, (addr), (cmd), (reqlen), (replylen))
+
+#defineIPMI_ALLOC_DRIVER_REQUEST(req, addr, cmd, reqlen, replylen)
\
+ (req) = __builtin_alloca(sizeof(struct ipmi_request) + \
+ (reqlen) + (replylen)); \
+ IPMI_INIT_DRIVER_REQUEST((req), (addr), (cmd), (reqlen),\
+ (replylen))
+
#ifdef IPMB
static int ipmi_ipmb_checksum(u_char, int);
static int ipmi_ipmb_send_message(device_t, u_char, u_char, u_char,
@@ -180,8 +197,8 @@ ipmi_dtor(void *arg)
*/
dev->ipmi_closing = 1;
while (dev->ipmi_requests > 0) {
- msleep(&dev->ipmi_requests, &sc->ipmi_lock, PWAIT,
- "ipmidrain", 0);
+ msleep(&dev->ipmi_requests, &sc->ipmi_requests_lock,
+ PWAIT, "ipmidrain", 0);
ipmi_purge_completed_requests(dev);
}
}
@@ -214,7 +231,7 @@ ipmi_ipmb_send_message(device_t dev, u_c
u_char slave_addr = 0x52;
int error;
- req = ipmi_alloc_driver_request(IPMI_ADDR(IPMI_APP_REQUEST, 0),
+ IPMI_ALLOC_DRIVER_REQUEST(req, IPMI_ADDR(IPMI_APP_REQUEST, 0),
IPMI_SEND_MSG, data_len + 8, 0);
req->ir_request[0] = channel;
req->ir_request[1] = slave_addr;
@@ -230,7 +247,6 @@ ipmi_ipmb_send_message(device_t dev, u_c
ipmi_submit_driver_request(sc, req);
error = req->ir_error;
- ipmi_free_request(req);
return (error);
}
@@ -242,7 +258,7 @@ ipmi_handle_attn(struct ipmi_softc *sc)
int error;
device_printf(sc->ipmi_dev, "BMC has a message\n");
- req = ipmi_alloc_driver_request(IPMI_ADDR(IPMI_APP_REQUEST, 0),
+ IPMI_ALLOC_DRIVER_REQUEST(req, IPMI_ADDR(IPMI_APP_REQUEST, 0),
IPMI_GET_MSG_FLAGS, 0, 1);
ipmi_submit_driver_request(sc, req);
@@ -256,9 +272,7 @@ ipmi_handle_attn(struct ipmi_softc *sc)
"watchdog about to go off");
}
if (req->ir_reply[0] & IPMI_MSG_AVAILABLE) {
- ipmi_free_request(req);
-
- req = ipmi_alloc_driver_request(
+ IPMI_ALLOC_DRIVER_REQUEST(req,
IPMI_ADDR(IPMI_APP_REQUEST, 0), IPMI_GET_MSG, 0,
16);
@@ -267,7 +281,6 @@ ipmi_handle_attn(struct ipmi_softc *sc)
}
}
error = req->ir_error;
- ipmi_free_request(req);
return (error);
}
@@ -477,15 +490,11 @@ ipmi_ioctl(struct cdev *cdev, u_long cmd
* Request management.
*/
-/* Allocate a new
svn commit: r279963 - in head/sys/dev: ahci ata mvs siis
Author: mav Date: Fri Mar 13 20:08:35 2015 New Revision: 279963 URL: https://svnweb.freebsd.org/changeset/base/279963 Log: Fix SATA Gen3 speed constants. MFC after:1 week Modified: head/sys/dev/ahci/ahci.h head/sys/dev/ata/ata-all.h head/sys/dev/mvs/mvs.h head/sys/dev/siis/siis.h Modified: head/sys/dev/ahci/ahci.h == --- head/sys/dev/ahci/ahci.hFri Mar 13 18:42:43 2015(r279962) +++ head/sys/dev/ahci/ahci.hFri Mar 13 20:08:35 2015(r279963) @@ -93,7 +93,7 @@ #define ATA_SS_SPD_NO_SPEED 0x #define ATA_SS_SPD_GEN1 0x0010 #define ATA_SS_SPD_GEN2 0x0020 -#define ATA_SS_SPD_GEN3 0x0040 +#define ATA_SS_SPD_GEN3 0x0030 #define ATA_SS_IPM_MASK 0x0f00 #define ATA_SS_IPM_NO_DEVICE0x @@ -131,7 +131,7 @@ #define ATA_SC_SPD_NO_SPEED 0x #define ATA_SC_SPD_SPEED_GEN1 0x0010 #define ATA_SC_SPD_SPEED_GEN2 0x0020 -#define ATA_SC_SPD_SPEED_GEN3 0x0040 +#define ATA_SC_SPD_SPEED_GEN3 0x0030 #define ATA_SC_IPM_MASK 0x0f00 #define ATA_SC_IPM_NONE 0x Modified: head/sys/dev/ata/ata-all.h == --- head/sys/dev/ata/ata-all.h Fri Mar 13 18:42:43 2015(r279962) +++ head/sys/dev/ata/ata-all.h Fri Mar 13 20:08:35 2015(r279963) @@ -105,6 +105,7 @@ #define ATA_SS_SPD_NO_SPEED 0x #define ATA_SS_SPD_GEN1 0x0010 #define ATA_SS_SPD_GEN2 0x0020 +#define ATA_SS_SPD_GEN3 0x0030 #define ATA_SS_IPM_MASK 0x0f00 #define ATA_SS_IPM_NO_DEVICE0x @@ -140,7 +141,7 @@ #define ATA_SC_SPD_NO_SPEED 0x #define ATA_SC_SPD_SPEED_GEN1 0x0010 #define ATA_SC_SPD_SPEED_GEN2 0x0020 -#define ATA_SC_SPD_SPEED_GEN3 0x0040 +#define ATA_SC_SPD_SPEED_GEN3 0x0030 #define ATA_SC_IPM_MASK 0x0f00 #define ATA_SC_IPM_NONE 0x Modified: head/sys/dev/mvs/mvs.h == --- head/sys/dev/mvs/mvs.h Fri Mar 13 18:42:43 2015(r279962) +++ head/sys/dev/mvs/mvs.h Fri Mar 13 20:08:35 2015(r279963) @@ -263,7 +263,7 @@ #defineSATA_SS_SPD_NO_SPEED 0x #defineSATA_SS_SPD_GEN1 0x0010 #defineSATA_SS_SPD_GEN2 0x0020 -#defineSATA_SS_SPD_GEN3 0x0040 +#defineSATA_SS_SPD_GEN3 0x0030 #defineSATA_SS_IPM_MASK 0x0f00 #defineSATA_SS_IPM_NO_DEVICE0x @@ -298,7 +298,7 @@ #defineSATA_SC_SPD_NO_SPEED 0x #defineSATA_SC_SPD_SPEED_GEN1 0x0010 #defineSATA_SC_SPD_SPEED_GEN2 0x0020 -#defineSATA_SC_SPD_SPEED_GEN3 0x0040 +#defineSATA_SC_SPD_SPEED_GEN3 0x0030 #defineSATA_SC_IPM_MASK 0x0f00 #defineSATA_SC_IPM_NONE 0x Modified: head/sys/dev/siis/siis.h == --- head/sys/dev/siis/siis.hFri Mar 13 18:42:43 2015(r279962) +++ head/sys/dev/siis/siis.hFri Mar 13 20:08:35 2015(r279963) @@ -92,7 +92,7 @@ #define ATA_SS_SPD_NO_SPEED 0x #define ATA_SS_SPD_GEN1 0x0010 #define ATA_SS_SPD_GEN2 0x0020 -#define ATA_SS_SPD_GEN3 0x0040 +#define ATA_SS_SPD_GEN3 0x0030 #define ATA_SS_IPM_MASK 0x0f00 #define ATA_SS_IPM_NO_DEVICE0x @@ -128,7 +128,7 @@ #define ATA_SC_SPD_NO_SPEED 0x #define ATA_SC_SPD_SPEED_GEN1 0x0010 #define ATA_SC_SPD_SPEED_GEN2 0x0020 -#define ATA_SC_SPD_SPEED_GEN3 0x0040 +#define ATA_SC_SPD_SPEED_GEN3 0x0030 #define ATA_SC_IPM_MASK 0x0f00 #define ATA_SC_IPM_NONE 0x ___ svn-src-all@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r279964 - in stable: 8/sys/dev/ipmi 9/sys/dev/ipmi
Author: jhb
Date: Fri Mar 13 20:10:09 2015
New Revision: 279964
URL: https://svnweb.freebsd.org/changeset/base/279964
Log:
MFC 278321:
Use direct hardware access for internal requests for KCS and SMIC. In
particular, updates to the watchdog should no longer sleep.
- Add a new IPMI_IO_LOCK for low-level I/O access. Use this for
kcs_polled_request() and smic_polled_request().
- Add a new backend callback "ipmi_driver_request" to handle a driver
request. The new callback performs the request sychronously for KCS
and SMIC. SSIF still defers the work to the worker thread since the
worker thread sleeps during request processing anyway.
- Allocate driver requests on the stack rather than using malloc().
Modified:
stable/9/sys/dev/ipmi/ipmi.c
stable/9/sys/dev/ipmi/ipmi_kcs.c
stable/9/sys/dev/ipmi/ipmi_smic.c
stable/9/sys/dev/ipmi/ipmi_ssif.c
stable/9/sys/dev/ipmi/ipmivars.h
Directory Properties:
stable/9/sys/ (props changed)
stable/9/sys/dev/ (props changed)
Changes in other areas also in this revision:
Modified:
stable/8/sys/dev/ipmi/ipmi.c
stable/8/sys/dev/ipmi/ipmi_kcs.c
stable/8/sys/dev/ipmi/ipmi_smic.c
stable/8/sys/dev/ipmi/ipmi_ssif.c
stable/8/sys/dev/ipmi/ipmivars.h
Directory Properties:
stable/8/sys/ (props changed)
stable/8/sys/dev/ (props changed)
stable/8/sys/dev/ipmi/ (props changed)
Modified: stable/9/sys/dev/ipmi/ipmi.c
==
--- stable/9/sys/dev/ipmi/ipmi.cFri Mar 13 20:08:35 2015
(r279963)
+++ stable/9/sys/dev/ipmi/ipmi.cFri Mar 13 20:10:09 2015
(r279964)
@@ -49,6 +49,23 @@ __FBSDID("$FreeBSD$");
#include
#endif
+/*
+ * Driver request structures are allocated on the stack via alloca() to
+ * avoid calling malloc(), especially for the watchdog handler.
+ * To avoid too much stack growth, a previously allocated structure can
+ * be reused via IPMI_INIT_DRIVER_REQUEST(), but the caller should ensure
+ * that there is adequate reply/request space in the original allocation.
+ */
+#defineIPMI_INIT_DRIVER_REQUEST(req, addr, cmd, reqlen, replylen)
\
+ bzero((req), sizeof(struct ipmi_request)); \
+ ipmi_init_request((req), NULL, 0, (addr), (cmd), (reqlen), (replylen))
+
+#defineIPMI_ALLOC_DRIVER_REQUEST(req, addr, cmd, reqlen, replylen)
\
+ (req) = __builtin_alloca(sizeof(struct ipmi_request) + \
+ (reqlen) + (replylen)); \
+ IPMI_INIT_DRIVER_REQUEST((req), (addr), (cmd), (reqlen),\
+ (replylen))
+
#ifdef IPMB
static int ipmi_ipmb_checksum(u_char, int);
static int ipmi_ipmb_send_message(device_t, u_char, u_char, u_char,
@@ -181,8 +198,8 @@ ipmi_dtor(void *arg)
*/
dev->ipmi_closing = 1;
while (dev->ipmi_requests > 0) {
- msleep(&dev->ipmi_requests, &sc->ipmi_lock, PWAIT,
- "ipmidrain", 0);
+ msleep(&dev->ipmi_requests, &sc->ipmi_requests_lock,
+ PWAIT, "ipmidrain", 0);
ipmi_purge_completed_requests(dev);
}
}
@@ -215,7 +232,7 @@ ipmi_ipmb_send_message(device_t dev, u_c
u_char slave_addr = 0x52;
int error;
- req = ipmi_alloc_driver_request(IPMI_ADDR(IPMI_APP_REQUEST, 0),
+ IPMI_ALLOC_DRIVER_REQUEST(req, IPMI_ADDR(IPMI_APP_REQUEST, 0),
IPMI_SEND_MSG, data_len + 8, 0);
req->ir_request[0] = channel;
req->ir_request[1] = slave_addr;
@@ -231,7 +248,6 @@ ipmi_ipmb_send_message(device_t dev, u_c
ipmi_submit_driver_request(sc, req);
error = req->ir_error;
- ipmi_free_request(req);
return (error);
}
@@ -243,7 +259,7 @@ ipmi_handle_attn(struct ipmi_softc *sc)
int error;
device_printf(sc->ipmi_dev, "BMC has a message\n");
- req = ipmi_alloc_driver_request(IPMI_ADDR(IPMI_APP_REQUEST, 0),
+ IPMI_ALLOC_DRIVER_REQUEST(req, IPMI_ADDR(IPMI_APP_REQUEST, 0),
IPMI_GET_MSG_FLAGS, 0, 1);
ipmi_submit_driver_request(sc, req);
@@ -257,9 +273,7 @@ ipmi_handle_attn(struct ipmi_softc *sc)
"watchdog about to go off");
}
if (req->ir_reply[0] & IPMI_MSG_AVAILABLE) {
- ipmi_free_request(req);
-
- req = ipmi_alloc_driver_request(
+ IPMI_ALLOC_DRIVER_REQUEST(req,
IPMI_ADDR(IPMI_APP_REQUEST, 0), IPMI_GET_MSG, 0,
16);
@@ -268,7 +282,6 @@ ipmi_handle_attn(struct ipmi_softc *sc)
}
}
error = req->ir_error;
- ipmi_free_request(req);
return (error);
}
@@ -478,15 +491,11 @@ ipmi_ioctl(struct cdev *cdev, u_long cmd
* Request management.
*/
-/* Allocate a new
svn commit: r279965 - head/usr.sbin/bhyve
Author: mav
Date: Fri Mar 13 20:14:35 2015
New Revision: 279965
URL: https://svnweb.freebsd.org/changeset/base/279965
Log:
Polish AHCI disk identify data and fix speed negotiation.
MFC after:2 weeks
Modified:
head/usr.sbin/bhyve/ahci.h
head/usr.sbin/bhyve/pci_ahci.c
Modified: head/usr.sbin/bhyve/ahci.h
==
--- head/usr.sbin/bhyve/ahci.h Fri Mar 13 20:10:09 2015(r279964)
+++ head/usr.sbin/bhyve/ahci.h Fri Mar 13 20:14:35 2015(r279965)
@@ -96,13 +96,14 @@
#define ATA_SS_SPD_NO_SPEED 0x
#define ATA_SS_SPD_GEN1 0x0010
#define ATA_SS_SPD_GEN2 0x0020
-#define ATA_SS_SPD_GEN3 0x0040
+#define ATA_SS_SPD_GEN3 0x0030
#define ATA_SS_IPM_MASK 0x0f00
#define ATA_SS_IPM_NO_DEVICE0x
#define ATA_SS_IPM_ACTIVE 0x0100
#define ATA_SS_IPM_PARTIAL 0x0200
#define ATA_SS_IPM_SLUMBER 0x0600
+#define ATA_SS_IPM_DEVSLEEP 0x0800
#define ATA_SERROR 14
#define ATA_SE_DATA_CORRECTED 0x0001
@@ -133,17 +134,19 @@
#define ATA_SC_SPD_NO_SPEED 0x
#define ATA_SC_SPD_SPEED_GEN1 0x0010
#define ATA_SC_SPD_SPEED_GEN2 0x0020
-#define ATA_SC_SPD_SPEED_GEN3 0x0040
+#define ATA_SC_SPD_SPEED_GEN3 0x0030
#define ATA_SC_IPM_MASK 0x0f00
#define ATA_SC_IPM_NONE 0x
#define ATA_SC_IPM_DIS_PARTIAL 0x0100
#define ATA_SC_IPM_DIS_SLUMBER 0x0200
+#define ATA_SC_IPM_DIS_DEVSLEEP 0x0400
#define ATA_SACTIVE 16
#define AHCI_MAX_PORTS 32
#define AHCI_MAX_SLOTS 32
+#define AHCI_MAX_IRQS 16
/* SATA AHCI v1.0 register defines */
#define AHCI_CAP0x00
@@ -208,6 +211,9 @@
#defineAHCI_CAP2_BOH 0x0001
#defineAHCI_CAP2_NVMP 0x0002
#defineAHCI_CAP2_APST 0x0004
+#defineAHCI_CAP2_SDS 0x0008
+#defineAHCI_CAP2_SADM 0x0010
+#defineAHCI_CAP2_DESO 0x0020
#define AHCI_OFFSET 0x100
#define AHCI_STEP 0x80
@@ -265,6 +271,7 @@
#define AHCI_P_CMD_ACTIVE 0x1000
#define AHCI_P_CMD_PARTIAL 0x2000
#define AHCI_P_CMD_SLUMBER 0x6000
+#define AHCI_P_CMD_DEVSLEEP 0x8000
#define AHCI_P_TFD 0x20
#define AHCI_P_SIG 0x24
@@ -284,6 +291,17 @@
#defineAHCI_P_FBS_ADO_SHIFT 12
#defineAHCI_P_FBS_DWE 0x000f
#defineAHCI_P_FBS_DWE_SHIFT 16
+#define AHCI_P_DEVSLP 0x44
+#defineAHCI_P_DEVSLP_ADSE 0x0001
+#defineAHCI_P_DEVSLP_DSP 0x0002
+#defineAHCI_P_DEVSLP_DETO 0x03fc
+#defineAHCI_P_DEVSLP_DETO_SHIFT 2
+#defineAHCI_P_DEVSLP_MDAT 0x7c00
+#defineAHCI_P_DEVSLP_MDAT_SHIFT 10
+#defineAHCI_P_DEVSLP_DITO 0x01ff8000
+#defineAHCI_P_DEVSLP_DITO_SHIFT 15
+#defineAHCI_P_DEVSLP_DM0x0e00
+#defineAHCI_P_DEVSLP_DM_SHIFT 25
/* Just to be sure, if building as module. */
#if MAXPHYS < 512 * 1024
Modified: head/usr.sbin/bhyve/pci_ahci.c
==
--- head/usr.sbin/bhyve/pci_ahci.c Fri Mar 13 20:10:09 2015
(r279964)
+++ head/usr.sbin/bhyve/pci_ahci.c Fri Mar 13 20:14:35 2015
(r279965)
@@ -431,7 +431,6 @@ ahci_port_stop(struct ahci_port *p)
static void
ahci_port_reset(struct ahci_port *pr)
{
- pr->sctl = 0;
pr->serr = 0;
pr->sact = 0;
pr->xfermode = ATA_UDMA6;
@@ -443,8 +442,11 @@ ahci_port_reset(struct ahci_port *pr)
pr->tfd = 0x7F;
return;
}
- pr->ssts = ATA_SS_DET_PHY_ONLINE | ATA_SS_SPD_GEN2 |
- ATA_SS_IPM_ACTIVE;
+ pr->ssts = ATA_SS_DET_PHY_ONLINE | ATA_SS_IPM_ACTIVE;
+ if (pr->sctl & ATA_SC_SPD_MASK)
+ pr->ssts |= (pr->sctl & ATA_SC_SPD_MASK);
+ else
+ pr->ssts |= ATA_SS_SPD_GEN3;
pr->tfd = (1 << 8) | ATA_S_DSC | ATA_S_DMA;
if (!pr->atapi) {
pr->sig = PxSIG_ATA;
@@ -470,6 +472,7 @@ ahci_reset(struct pci_ahci_softc *sc)
for (i = 0; i < sc->ports; i++) {
sc->port[i].ie = 0;
sc->port[i].is = 0;
+ sc->port[i].sctl = 0;
ahci_port_reset(&sc->port[i]);
}
}
@@ -807,26 +810,36 @@ handle_identify(struct ahci_port *p, int
buf[53] = (1 << 1 | 1 << 2);
if (p->mult_sectors)
buf[59] = (0x100 | p->mult_secto
svn commit: r279966 - head/contrib/libxo/libxo
Author: trasz Date: Fri Mar 13 20:14:55 2015 New Revision: 279966 URL: https://svnweb.freebsd.org/changeset/base/279966 Log: Mark xo_err(3), xo_errx(3), and xo_errc(3) as __dead2. Differential Revision:https://reviews.freebsd.org/D2059 Reviewed by: marcel@ MFC after:1 month Sponsored by: The FreeBSD Foundation Modified: head/contrib/libxo/libxo/xo.h Modified: head/contrib/libxo/libxo/xo.h == --- head/contrib/libxo/libxo/xo.h Fri Mar 13 20:14:35 2015 (r279965) +++ head/contrib/libxo/libxo/xo.h Fri Mar 13 20:14:55 2015 (r279966) @@ -272,13 +272,13 @@ void xo_warnx (const char *fmt, ...); void -xo_err (int eval, const char *fmt, ...); +xo_err (int eval, const char *fmt, ...) __dead2; void -xo_errx (int eval, const char *fmt, ...); +xo_errx (int eval, const char *fmt, ...) __dead2; void -xo_errc (int eval, int code, const char *fmt, ...); +xo_errc (int eval, int code, const char *fmt, ...) __dead2; void xo_message_hcv (xo_handle_t *xop, int code, const char *fmt, va_list vap); ___ svn-src-all@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
Re: svn commit: r279932 - head/sys/vm
On Fri, 2015-03-13 at 14:34 -0400, John Baldwin wrote:
> On Friday, March 13, 2015 11:57:58 AM Ian Lepore wrote:
> > On Fri, 2015-03-13 at 13:19 -0400, John Baldwin wrote:
> > > On Friday, March 13, 2015 10:14:27 AM Ian Lepore wrote:
> > > > On Fri, 2015-03-13 at 06:24 -0400, John Baldwin wrote:
> > > > > On Thursday, March 12, 2015 05:24:51 PM Ian Lepore wrote:
> > [...]
> > > >
> > > > In general I'm glad I got called away to an onsite meeting yesterday and
> > > > didn't get far with these changes, because the more I think about it,
> > > > the less satisfied I am with this expedient fix. The other fix I
> > > > started on, where a new SBUF_COUNTNUL flag can be set to inform the
> > > > sbuf_finish() code that you want the terminating nul counted in the data
> > > > length just feels like a better fit for the overall "automaticness" of
> > > > how the sbuf stuff works.
> > >
> > > Hmm, I actually think that it's a bug that the terminating nul isn't
> > > included
> > > when draining. If we fixed that then I think that fixes most of these?
> > > The places that explicitly use 'sysctl_handle_string()' with an sbuf
> > > should probably just be using sbuf_len(sb) + 1' explicitly. (Another
> > > option would be to have a sysctl_handle_sbuf() that was a wrapper around
> > > sysctl_handle_string() that included the + 1 to hide that detail if there
> > > is
> > > more than one.)
> > >
> >
> > Some of the uses of sbuf for sysctl use sbuf_bcat() for dealing with
> > binary structs, so we can't just assume that a nullterm should be added
> > and included in the buffer length -- there needs to be some mechanism to
> > say explicitly "this is an sbuf for a sysctl string" (and more generally
> > "this is an sbuf for a string where I want the nul byte counted as part
> > of the data" because that could be useful in non-sysctl contexts too,
> > especially in userland).
>
> Humm, that would seem to be an abuse of the API really. It is specifically
> designed for strings as someone else noted at the start of this thread (and
> as noted in the manpage). If anything I'd argue that the use cases that don't
> want a string should be the ones that should get a special flag in that case
> (or perhaps we should have a different little API to manage a buffer used for
> a draining sysctl where the data is a binary blob instead of a string). If
> you agree I'm happy to do some of the work (e.g. the different wrapper API).
>
Given the existance of sbuf_bcpy() and sbuf_bcat() I'm not sure we can
say using sbuf for binary data is any kind of violation; somebody just
used the API that was provided to solve their problem.
Binary data is the exception in the sysctl case, and the idea of having
sbuf_new_for_sysctl() assume you're setting up to handle a sysctl string
and requiring the rare binary uses to do something different does make a
lot of sense. That might lead to a patch like the one below, which
would automatically fix most of the current sysctl sbuf users, and the 2
or 3 places that are using binary data would need to add a line:
sbuf_clear_flags(&sbuf, SBUF_INCLUDENUL);
I should mention too that the larger problem I'm trying to clean up is
that some sysctl strings include the nul byte in the data returned to
userland and some don't. There are more direct callers of SYSCTL_OUT()
that fail to add a nulterm, I have a whole separate set of fixes for
those, but I'm becoming somewhat inclined to fix them by converting them
to use sbuf and just make that the established idiom for returning
dynamic strings via sysctl.
-- Ian
Index: sys/kern/kern_sysctl.c
===
--- sys/kern/kern_sysctl.c (revision 279962)
+++ sys/kern/kern_sysctl.c (working copy)
@@ -1807,7 +1807,7 @@ sbuf_new_for_sysctl(struct sbuf *s, char *buf, int
struct sysctl_req *req)
{
- s = sbuf_new(s, buf, length, SBUF_FIXEDLEN);
+ s = sbuf_new(s, buf, length, SBUF_FIXEDLEN | SBUF_INCLUDENUL);
sbuf_set_drain(s, sbuf_sysctl_drain, req);
return (s);
}
Index: sys/kern/subr_sbuf.c
===
--- sys/kern/subr_sbuf.c (revision 279962)
+++ sys/kern/subr_sbuf.c (working copy)
@@ -262,6 +262,28 @@ sbuf_uionew(struct sbuf *s, struct uio *uio, int *
}
#endif
+int
+sbuf_get_flags(struct sbuf *s)
+{
+
+ return (s->s_flags);
+}
+
+void
+sbuf_clear_flags(struct sbuf *s, int flags)
+{
+
+ s->s_flags &= ~(flags & SBUF_USRFLAGMSK);
+}
+
+void
+sbuf_set_flags(struct sbuf *s, int flags)
+{
+
+
+ s->s_flags |= (flags & SBUF_USRFLAGMSK);
+}
+
/*
* Clear an sbuf and reset its position.
*/
@@ -697,11 +719,13 @@ sbuf_finish(struct sbuf *s)
assert_sbuf_integrity(s);
assert_sbuf_state(s, 0);
+ s->s_buf[s->s_len] = '\0';
+ if (s->s_flags & SBUF_INCLUDENUL)
+ s->s_len++;
if (s->s_drain_func != NULL) {
while (s->s_len > 0 && s->s_error == 0)
s->s_error = sbuf_drain(s);
}
- s->s_buf[s->s_len] = '\0';
SBUF_SETFLAG(s, SBUF_
svn commit: r279967 - head/usr.sbin/bhyve
Author: mav
Date: Fri Mar 13 20:56:17 2015
New Revision: 279967
URL: https://svnweb.freebsd.org/changeset/base/279967
Log:
Change prdbc value reporting.
MFC after:2 weeks
Modified:
head/usr.sbin/bhyve/pci_ahci.c
Modified: head/usr.sbin/bhyve/pci_ahci.c
==
--- head/usr.sbin/bhyve/pci_ahci.c Fri Mar 13 20:14:55 2015
(r279966)
+++ head/usr.sbin/bhyve/pci_ahci.c Fri Mar 13 20:56:17 2015
(r279967)
@@ -1662,6 +1662,9 @@ ata_ioreq_cb(struct blockif_req *br, int
*/
STAILQ_INSERT_TAIL(&p->iofhd, aior, io_flist);
+ if (!err)
+ hdr->prdbc = aior->done;
+
if (dsm) {
if (aior->done != aior->len && !err) {
ahci_handle_dsm_trim(p, slot, cfis, aior->done);
@@ -1677,13 +1680,8 @@ ata_ioreq_cb(struct blockif_req *br, int
if (!err && aior->done == aior->len) {
tfd = ATA_S_READY | ATA_S_DSC;
- if (ncq)
- hdr->prdbc = 0;
- else
- hdr->prdbc = aior->len;
} else {
tfd = (ATA_E_ABORT << 8) | ATA_S_READY | ATA_S_ERROR;
- hdr->prdbc = 0;
if (ncq)
p->serr |= (1 << slot);
}
@@ -1738,6 +1736,9 @@ atapi_ioreq_cb(struct blockif_req *br, i
*/
STAILQ_INSERT_TAIL(&p->iofhd, aior, io_flist);
+ if (!err)
+ hdr->prdbc = aior->done;
+
if (pending && !err) {
atapi_read(p, slot, cfis, aior->done, hdr->prdtl - pending);
goto out;
@@ -1745,12 +1746,10 @@ atapi_ioreq_cb(struct blockif_req *br, i
if (!err && aior->done == aior->len) {
tfd = ATA_S_READY | ATA_S_DSC;
- hdr->prdbc = aior->len;
} else {
p->sense_key = ATA_SENSE_ILLEGAL_REQUEST;
p->asc = 0x21;
tfd = (p->sense_key << 12) | ATA_S_READY | ATA_S_ERROR;
- hdr->prdbc = 0;
}
cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
___
svn-src-all@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/svn-src-all
To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r279968 - head/usr.sbin/bhyve
Author: mav
Date: Fri Mar 13 21:01:25 2015
New Revision: 279968
URL: https://svnweb.freebsd.org/changeset/base/279968
Log:
Remove incorrect SERR register setting.
At this point we have nothing to report through that register.
MFC after:2 weeks
Modified:
head/usr.sbin/bhyve/pci_ahci.c
Modified: head/usr.sbin/bhyve/pci_ahci.c
==
--- head/usr.sbin/bhyve/pci_ahci.c Fri Mar 13 20:56:17 2015
(r279967)
+++ head/usr.sbin/bhyve/pci_ahci.c Fri Mar 13 21:01:25 2015
(r279968)
@@ -1682,8 +1682,6 @@ ata_ioreq_cb(struct blockif_req *br, int
tfd = ATA_S_READY | ATA_S_DSC;
} else {
tfd = (ATA_E_ABORT << 8) | ATA_S_READY | ATA_S_ERROR;
- if (ncq)
- p->serr |= (1 << slot);
}
if (ncq) {
___
svn-src-all@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/svn-src-all
To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
Re: svn commit: r279965 - head/usr.sbin/bhyve
On Friday, March 13, 2015 08:14:35 PM Alexander Motin wrote: > Author: mav > Date: Fri Mar 13 20:14:35 2015 > New Revision: 279965 > URL: https://svnweb.freebsd.org/changeset/base/279965 > > Log: > Polish AHCI disk identify data and fix speed negotiation. > > MFC after: 2 weeks > > Modified: > head/usr.sbin/bhyve/ahci.h > head/usr.sbin/bhyve/pci_ahci.c > > Modified: head/usr.sbin/bhyve/ahci.h > == > --- head/usr.sbin/bhyve/ahci.hFri Mar 13 20:10:09 2015 > (r279964) > +++ head/usr.sbin/bhyve/ahci.hFri Mar 13 20:14:35 2015 > (r279965) > @@ -96,13 +96,14 @@ > #define ATA_SS_SPD_NO_SPEED 0x > #define ATA_SS_SPD_GEN1 0x0010 > #define ATA_SS_SPD_GEN2 0x0020 > -#define ATA_SS_SPD_GEN3 0x0040 > +#define ATA_SS_SPD_GEN3 0x0030 > > #define ATA_SS_IPM_MASK 0x0f00 > #define ATA_SS_IPM_NO_DEVICE0x > #define ATA_SS_IPM_ACTIVE 0x0100 > #define ATA_SS_IPM_PARTIAL 0x0200 > #define ATA_SS_IPM_SLUMBER 0x0600 > +#define ATA_SS_IPM_DEVSLEEP 0x0800 I wonder if it would make sense to have a shared header for these constants (given that it seems they are copied in 4 other places for in-kernel drivers)? -- John Baldwin ___ svn-src-all@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
Re: svn commit: r279932 - head/sys/vm
On Friday, March 13, 2015 02:42:13 PM Ian Lepore wrote: > On Fri, 2015-03-13 at 14:34 -0400, John Baldwin wrote: > > On Friday, March 13, 2015 11:57:58 AM Ian Lepore wrote: > > > On Fri, 2015-03-13 at 13:19 -0400, John Baldwin wrote: > > > > On Friday, March 13, 2015 10:14:27 AM Ian Lepore wrote: > > > > > On Fri, 2015-03-13 at 06:24 -0400, John Baldwin wrote: > > > > > > On Thursday, March 12, 2015 05:24:51 PM Ian Lepore wrote: > > > [...] > > > > > > > > > > In general I'm glad I got called away to an onsite meeting yesterday > > > > > and > > > > > didn't get far with these changes, because the more I think about it, > > > > > the less satisfied I am with this expedient fix. The other fix I > > > > > started on, where a new SBUF_COUNTNUL flag can be set to inform the > > > > > sbuf_finish() code that you want the terminating nul counted in the > > > > > data > > > > > length just feels like a better fit for the overall "automaticness" of > > > > > how the sbuf stuff works. > > > > > > > > Hmm, I actually think that it's a bug that the terminating nul isn't > > > > included > > > > when draining. If we fixed that then I think that fixes most of these? > > > > The places that explicitly use 'sysctl_handle_string()' with an sbuf > > > > should probably just be using sbuf_len(sb) + 1' explicitly. (Another > > > > option would be to have a sysctl_handle_sbuf() that was a wrapper around > > > > sysctl_handle_string() that included the + 1 to hide that detail if > > > > there is > > > > more than one.) > > > > > > > > > > Some of the uses of sbuf for sysctl use sbuf_bcat() for dealing with > > > binary structs, so we can't just assume that a nullterm should be added > > > and included in the buffer length -- there needs to be some mechanism to > > > say explicitly "this is an sbuf for a sysctl string" (and more generally > > > "this is an sbuf for a string where I want the nul byte counted as part > > > of the data" because that could be useful in non-sysctl contexts too, > > > especially in userland). > > > > Humm, that would seem to be an abuse of the API really. It is specifically > > designed for strings as someone else noted at the start of this thread (and > > as noted in the manpage). If anything I'd argue that the use cases that > > don't > > want a string should be the ones that should get a special flag in that case > > (or perhaps we should have a different little API to manage a buffer used > > for > > a draining sysctl where the data is a binary blob instead of a string). If > > you agree I'm happy to do some of the work (e.g. the different wrapper API). > > > > Given the existance of sbuf_bcpy() and sbuf_bcat() I'm not sure we can > say using sbuf for binary data is any kind of violation; somebody just > used the API that was provided to solve their problem. Well, it still nul-terminates the result of those, so I think it's still really dealing with strings. Those can be useful for appending non nul-terminated strings (for example using the d_namelen from a dirent). However, I think an INCLUDENUL flag is fine. It is a smaller change than adding a new sysctl-drain API. > Binary data is the exception in the sysctl case, and the idea of having > sbuf_new_for_sysctl() assume you're setting up to handle a sysctl string > and requiring the rare binary uses to do something different does make a > lot of sense. That might lead to a patch like the one below, which > would automatically fix most of the current sysctl sbuf users, and the 2 > or 3 places that are using binary data would need to add a line: > >sbuf_clear_flags(&sbuf, SBUF_INCLUDENUL); > > I should mention too that the larger problem I'm trying to clean up is > that some sysctl strings include the nul byte in the data returned to > userland and some don't. There are more direct callers of SYSCTL_OUT() > that fail to add a nulterm, I have a whole separate set of fixes for > those, but I'm becoming somewhat inclined to fix them by converting them > to use sbuf and just make that the established idiom for returning > dynamic strings via sysctl. Either that or using sysctl_handle_string() when possible instead of bare SYSCTL_OUT(). I think your patch is fine. I agree this will be a much smaller change to roll out. :) -- John Baldwin ___ svn-src-all@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/svn-src-all To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
Re: svn commit: r279937 - in head/sys/powerpc: include powerpc
On Fri, 13 Mar 2015, Konstantin Belousov wrote: On Fri, Mar 13, 2015 at 10:45:46AM -0700, Nathan Whitehorn wrote: We'll need to hack the compiler in this case, since it assumes setjmp() saves and restores the vector registers. I'm really not sure which option is worse. Changing the compiler is arguably much worse than breaking ABI of the tier 2 platform, indeed. We must maintain the situation where the stock build of the compilers work out of box. Still, how the compiler' assumptions are laid out ? It could be argued that compilers on x86 also assume that FPU register file is restored by longjmp. That would be a compiler bug. %st* and %xmm* are defined as not preserved across function calls, but I suspect that practical rule for setjmp() is that floating vars better not be used in the target frame. Indeed. They are not call-saved for ordinary functions, and it would be stupid for the compiler to treat setjmp() differently, since setjmp() wants to clobber the registers even more than an ordinary functions. Hm, indeed f14-f31 and v20-v31 are marked as non-volatile for 64bit ABI. Powerpc has zillions of call-saved registers (I know little about powerpc, but checked this in its gcc .md file). There are complications for the FP environment. setjmp() is specified to save "its" environoment and longjmp() is specified to restore the environment saved by setjmp(). Environments seem to be under-specified. Floating point has an state that is actually called an environment, but is this part of the execution environment? Anyway, it is clear that most call-saved parts of the FP environment should be saved and restored (so that longjmp from signal handlers has a chance of working), but the exception flags should not be restored (since restoring them would forget exceptions). x86 setjmp/longjump attempts to do this, but is still very buggy: - i386 _setjmp saves the i387 control word but not not the SSE control word (if any). This is the setjmp that doesn't save the signal mask. - i386 ___longjmp cannot restore the SSE control word since it was not saved. (_longjmp is a weak alias for ___longjmp. I don't know why _longjmp needs aliasing but _setjmp doesn't. Similarly for the other 5 *setjmp/*longjmp pairs in x86.) - to preserve (that is, _not_ restore) the exception flags, someone removed the fninit that I wrote in all the i386 longjmps. This broke: - cleaning up of the state. Especially the tag word. Cleaning is needed for at least longjmp() from FreeBSD-[1-4]-compat signal handlers which can be entered with any FP state - avoiding getting an exception when reloading the FP control word. The correct change is something like: fnstsw %ax fninit fstenv tmp(%esp) # set the pending-exception bit in %ax if there are any unmasked # exceptions for the _new_ cw, else clear it # store %ax in tmp env # copy saved cw from jmp_buf to tmp env fldenv tmp(%esp) # do related stuff for SSE (simpler) The trip through the tmp env is needed since there is no instruction to load the new status word from a registers; anyway, it would be difficult to load the cw and status word separately without risking a trap for one (fninit only cleans enough for the next instruction). - i386 setjmp and __longjmp are identically broken. - i386 sigsetjmp and __siglongjmp are not even identically broken. They used to be almost identical copies of setjmp and __longjmp (sigsetjmp == setjmp in BSD). They shouldn't exist except as strong aliases, but still do despite extensive use of weak aliases and namespace messes like the triply underscored ___longjmp. No one removed the fninit's from them. So they are missing most of the above bugs, but the exception flags are not preserved as intended. - amd64 _setjmp, setjmp, and sigsetjmp are OK, except for the existence of sigsetjmp as a non-alias. They do preserve mxcsr. The bug of the existence of the file containing sigsetjmp was copied perfectly from i386, including my comments in it saying why it should not exist. - amd64 ___longjmp does preserve SSE exceptions. This requires merging the current exceptions with the saved control word in mxcsr. SSE exceptions are synchronous, and ldmxcsr cannot trap, so I think it is not necessary to adjust the pending-exception flag (or there is no such flag in mxcsr). amd64 _longjmp then has the same bugs as i386 _longjmp restoring the i387 cw, tag word and pending-exception flag without trapping. - amd64 __longjmp is like amd64 ___longjmp. - amd64 __siglongjmp is not even like amd64 ___longjmp. No one removed the fninit there either. These details are of course arcane and rarely matter. I tested the old versions using longjmp from signal handlers. FreeBSD-[1-4] signal handlers get the interrupted FP context. One interesting case is when the compiler temporarily changes the rounding mode. This may be interrupted by even a non-FP relate
svn commit: r279969 - head/sys/dev/cxgbe
Author: np
Date: Sat Mar 14 00:02:53 2015
New Revision: 279969
URL: https://svnweb.freebsd.org/changeset/base/279969
Log:
cxgbe(4): fix if_media handling for T520-BT cards. 1Gbps and 100Mbps
are valid for this card.
MFC after:1 week
Modified:
head/sys/dev/cxgbe/t4_main.c
Modified: head/sys/dev/cxgbe/t4_main.c
==
--- head/sys/dev/cxgbe/t4_main.cFri Mar 13 21:01:25 2015
(r279968)
+++ head/sys/dev/cxgbe/t4_main.cSat Mar 14 00:02:53 2015
(r279969)
@@ -2895,9 +2895,6 @@ build_medialist(struct port_info *pi, st
switch(pi->port_type) {
case FW_PORT_TYPE_BT_XFI:
- ifmedia_add(media, m | IFM_10G_T, data, NULL);
- break;
-
case FW_PORT_TYPE_BT_XAUI:
ifmedia_add(media, m | IFM_10G_T, data, NULL);
/* fall through */
___
svn-src-all@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/svn-src-all
To unsubscribe, send any mail to "svn-src-all-unsubscr...@freebsd.org"
svn commit: r279970 - in head/sys/x86: include x86 xen
Author: neel
Date: Sat Mar 14 00:30:41 2015
New Revision: 279970
URL: https://svnweb.freebsd.org/changeset/base/279970
Log:
Add x86 specific APIs 'lapic_ipi_alloc()' and 'lapic_ipi_free()' to allow IPI
vectors to be dynamically allocated. This allows kernel modules like vmm.ko
to allocate unique IPI slots when loaded (as opposed to hard allocating one
or more vectors).
Also, reorganize the fixed IPI vectors to create a contiguous space for
dynamic IPI allocation.
Reviewed by: kib, jhb
Differential Revision:https://reviews.freebsd.org/D2042
Modified:
head/sys/x86/include/apicvar.h
head/sys/x86/x86/local_apic.c
head/sys/x86/xen/xen_apic.c
Modified: head/sys/x86/include/apicvar.h
==
--- head/sys/x86/include/apicvar.h Sat Mar 14 00:02:53 2015
(r279969)
+++ head/sys/x86/include/apicvar.h Sat Mar 14 00:30:41 2015
(r279970)
@@ -111,11 +111,8 @@
#defineIPI_INVLPG (APIC_IPI_INTS + 2)
#defineIPI_INVLRNG (APIC_IPI_INTS + 3)
#defineIPI_INVLCACHE (APIC_IPI_INTS + 4)
-#ifdef __i386__
-#defineIPI_LAZYPMAP(APIC_IPI_INTS + 5) /* Lazy pmap release. */
-#endif
/* Vector to handle bitmap based IPIs */
-#defineIPI_BITMAP_VECTOR (APIC_IPI_INTS + 6)
+#defineIPI_BITMAP_VECTOR (APIC_IPI_INTS + 5)
/* IPIs handled by IPI_BITMAP_VECTOR */
#defineIPI_AST 0 /* Generate software trap. */
@@ -124,8 +121,15 @@
#define IPI_BITMAP_LAST IPI_HARDCLOCK
#define IPI_IS_BITMAPED(x) ((x) <= IPI_BITMAP_LAST)
-#defineIPI_STOP(APIC_IPI_INTS + 7) /* Stop CPU until
restarted. */
-#defineIPI_SUSPEND (APIC_IPI_INTS + 8) /* Suspend CPU until
restarted. */
+#defineIPI_STOP(APIC_IPI_INTS + 6) /* Stop CPU until
restarted. */
+#defineIPI_SUSPEND (APIC_IPI_INTS + 7) /* Suspend CPU until
restarted. */
+#ifdef __i386__
+#defineIPI_LAZYPMAP(APIC_IPI_INTS + 8) /* Lazy pmap release. */
+#defineIPI_DYN_FIRST (APIC_IPI_INTS + 9)
+#else
+#defineIPI_DYN_FIRST (APIC_IPI_INTS + 8)
+#endif
+#defineIPI_DYN_LAST(254) /* IPIs allocated at
runtime */
/*
* IPI_STOP_HARD does not need to occupy a slot in the IPI vector space since
@@ -224,6 +228,8 @@ struct apic_ops {
void(*ipi_raw)(register_t, u_int);
void(*ipi_vectored)(u_int, int);
int (*ipi_wait)(int);
+ int (*ipi_alloc)(inthand_t *ipifunc);
+ void(*ipi_free)(int vector);
/* LVT */
int (*set_lvt_mask)(u_int, u_int, u_char);
@@ -397,6 +403,20 @@ lapic_ipi_wait(int delay)
}
static inline int
+lapic_ipi_alloc(inthand_t *ipifunc)
+{
+
+ return (apic_ops.ipi_alloc(ipifunc));
+}
+
+static inline void
+lapic_ipi_free(int vector)
+{
+
+ return (apic_ops.ipi_free(vector));
+}
+
+static inline int
lapic_set_lvt_mask(u_int apic_id, u_int lvt, u_char masked)
{
Modified: head/sys/x86/x86/local_apic.c
==
--- head/sys/x86/x86/local_apic.c Sat Mar 14 00:02:53 2015
(r279969)
+++ head/sys/x86/x86/local_apic.c Sat Mar 14 00:30:41 2015
(r279970)
@@ -303,6 +303,8 @@ static int native_lapic_set_lvt_polarit
enum intr_polarity pol);
static int native_lapic_set_lvt_triggermode(u_int apic_id, u_int lvt,
enum intr_trigger trigger);
+static int native_lapic_ipi_alloc(inthand_t *ipifunc);
+static voidnative_lapic_ipi_free(int vector);
struct apic_ops apic_ops = {
.create = native_lapic_create,
@@ -329,6 +331,8 @@ struct apic_ops apic_ops = {
.ipi_raw= native_lapic_ipi_raw,
.ipi_vectored = native_lapic_ipi_vectored,
.ipi_wait = native_lapic_ipi_wait,
+ .ipi_alloc = native_lapic_ipi_alloc,
+ .ipi_free = native_lapic_ipi_free,
#endif
.set_lvt_mask = native_lapic_set_lvt_mask,
.set_lvt_mode = native_lapic_set_lvt_mode,
@@ -1761,4 +1765,60 @@ native_lapic_ipi_vectored(u_int vector,
}
#endif /* DETECT_DEADLOCK */
}
+
+/*
+ * Since the IDT is shared by all CPUs the IPI slot update needs to be globally
+ * visible.
+ *
+ * Consider the case where an IPI is generated immediately after allocation:
+ * vector = lapic_ipi_alloc(ipifunc);
+ * ipi_selected(other_cpus, vector);
+ *
+ * In xAPIC mode a write to ICR_LO has serializing semantics because the
+ * APIC page is mapped as an uncached region. In x2APIC mode there is an
+ * explicit 'mfence' before the ICR MSR is written. Therefore in both cases
+ * the IDT slot update is globally visible before the IPI is delivered.
+ */
+static int
+native_lapic_ipi_alloc(inthand_t *ipifunc)
+
svn commit: r279971 - in head/sys: amd64/amd64 amd64/include amd64/vmm amd64/vmm/intel amd64/vmm/io modules/vmm
Author: neel
Date: Sat Mar 14 02:32:08 2015
New Revision: 279971
URL: https://svnweb.freebsd.org/changeset/base/279971
Log:
Use lapic_ipi_alloc() to dynamically allocate IPI slots needed by bhyve when
vmm.ko is loaded.
Also relocate the 'justreturn' IPI handler to be alongside all other handlers.
Requested by: kib
Deleted:
head/sys/amd64/vmm/vmm_ipi.c
head/sys/amd64/vmm/vmm_ipi.h
head/sys/amd64/vmm/vmm_support.S
Modified:
head/sys/amd64/amd64/apic_vector.S
head/sys/amd64/include/smp.h
head/sys/amd64/vmm/intel/ept.c
head/sys/amd64/vmm/intel/vmx.c
head/sys/amd64/vmm/io/vlapic.c
head/sys/amd64/vmm/vmm.c
head/sys/amd64/vmm/vmm_lapic.c
head/sys/modules/vmm/Makefile
Modified: head/sys/amd64/amd64/apic_vector.S
==
--- head/sys/amd64/amd64/apic_vector.S Sat Mar 14 00:30:41 2015
(r279970)
+++ head/sys/amd64/amd64/apic_vector.S Sat Mar 14 02:32:08 2015
(r279971)
@@ -301,4 +301,35 @@ IDTVEC(rendezvous)
callsmp_rendezvous_action
callas_lapic_eoi
jmp doreti
+
+/*
+ * IPI handler whose purpose is to interrupt the CPU with minimum overhead.
+ * This is used by bhyve to force a host cpu executing in guest context to
+ * trap into the hypervisor.
+ *
+ * This handler is different from other IPI handlers in the following aspects:
+ *
+ * 1. It doesn't push a trapframe on the stack.
+ *
+ * This implies that a DDB backtrace involving 'justreturn' will skip the
+ * function that was interrupted by this handler.
+ *
+ * 2. It doesn't 'swapgs' when userspace is interrupted.
+ *
+ * The 'justreturn' handler does not access any pcpu data so it is not an
+ * issue. Moreover the 'justreturn' handler can only be interrupted by an NMI
+ * whose handler already doesn't trust GS.base when kernel code is interrupted.
+ */
+ .text
+ SUPERALIGN_TEXT
+IDTVEC(justreturn)
+ pushq %rax
+ pushq %rcx
+ pushq %rdx
+ callas_lapic_eoi
+ popq%rdx
+ popq%rcx
+ popq%rax
+ jmp doreti_iret
+
#endif /* SMP */
Modified: head/sys/amd64/include/smp.h
==
--- head/sys/amd64/include/smp.hSat Mar 14 00:30:41 2015
(r279970)
+++ head/sys/amd64/include/smp.hSat Mar 14 02:32:08 2015
(r279971)
@@ -54,6 +54,7 @@ inthand_t
IDTVEC(ipi_intr_bitmap_handler), /* Bitmap based IPIs */
IDTVEC(cpustop),/* CPU stops & waits to be restarted */
IDTVEC(cpususpend), /* CPU suspends & waits to be resumed */
+ IDTVEC(justreturn), /* interrupt CPU with minimum overhead */
IDTVEC(rendezvous); /* handle CPU rendezvous */
struct pmap;
Modified: head/sys/amd64/vmm/intel/ept.c
==
--- head/sys/amd64/vmm/intel/ept.c Sat Mar 14 00:30:41 2015
(r279970)
+++ head/sys/amd64/vmm/intel/ept.c Sat Mar 14 02:32:08 2015
(r279971)
@@ -43,7 +43,6 @@ __FBSDID("$FreeBSD$");
#include
#include "vmx_cpufunc.h"
-#include "vmm_ipi.h"
#include "ept.h"
#defineEPT_SUPPORTS_EXEC_ONLY(cap) ((cap) & (1UL << 0))
Modified: head/sys/amd64/vmm/intel/vmx.c
==
--- head/sys/amd64/vmm/intel/vmx.c Sat Mar 14 00:30:41 2015
(r279970)
+++ head/sys/amd64/vmm/intel/vmx.c Sat Mar 14 02:32:08 2015
(r279971)
@@ -55,7 +55,6 @@ __FBSDID("$FreeBSD$");
#include "vmm_lapic.h"
#include "vmm_host.h"
#include "vmm_ioport.h"
-#include "vmm_ipi.h"
#include "vmm_ktr.h"
#include "vmm_stat.h"
#include "vatpic.h"
@@ -175,7 +174,7 @@ static int posted_interrupts;
SYSCTL_INT(_hw_vmm_vmx_cap, OID_AUTO, posted_interrupts, CTLFLAG_RD,
&posted_interrupts, 0, "APICv posted interrupt support");
-static int pirvec;
+static int pirvec = -1;
SYSCTL_INT(_hw_vmm_vmx, OID_AUTO, posted_interrupt_vector, CTLFLAG_RD,
&pirvec, 0, "APICv posted interrupt vector");
@@ -485,8 +484,8 @@ static int
vmx_cleanup(void)
{
- if (pirvec != 0)
- vmm_ipi_free(pirvec);
+ if (pirvec >= 0)
+ lapic_ipi_free(pirvec);
if (vpid_unr != NULL) {
delete_unrhdr(vpid_unr);
@@ -694,8 +693,8 @@ vmx_init(int ipinum)
MSR_VMX_TRUE_PINBASED_CTLS, PINBASED_POSTED_INTERRUPT, 0,
&tmp);
if (error == 0) {
- pirvec = vmm_ipi_alloc();
- if (pirvec == 0) {
+ pirvec = lapic_ipi_alloc(&IDTVEC(justreturn));
+ if (pirvec < 0) {
if (bootverbose) {
printf("vmx_init: unable to allocate "
"posted interrupt vector\n")
