oot option; they are always on now
How long have they been present for?
Wouldn't it be best to leave them as valid parameters??
David
--
David Laight: da...@l8s.co.uk
On Mon, Oct 01, 2012 at 12:48:24AM +, David Holland wrote:
> On Sun, Sep 30, 2012 at 08:19:52PM +0000, David Laight wrote:
> > Unfortunately apmbios.h is used by the world !
>
> How much of it, and what would happen if we removed it entirely?
> (Also, what's the recom
nux guys are busy adding new BPF codes?
David
--
David Laight: da...@l8s.co.uk
st.c
>
> Log Message:
> Remove sljit because it was imported to the wrong place.
How long was it there? and was it ever in a build?
If not built it can be deleted from the repository.
David
--
David Laight: da...@l8s.co.uk
saligned.
I suspect that the 'sub %sp,n' use to allocate a stack frame has
hardware optimisations that are absent from the 'and %sp,~7' needed
to realign it.
David
--
David Laight: da...@l8s.co.uk
deal with "issues" of this interface.
> Seems it can't do DMA updates of the rxsts for mbufs with addresses
> >= 256MB. It can't also seem to properly read from the descriptor rings
> if they are < 256MB. And I have no idea why either should matter.
The combination
ild inodes don't even need looking at.
I suspect the DNLC (and any other caches) will need both rename ops done
atomically though.
David
--
David Laight: da...@l8s.co.uk
On Sat, Oct 20, 2012 at 04:06:13AM +, David Holland wrote:
> On Fri, Oct 19, 2012 at 07:58:34AM +0100, David Laight wrote:
> > > Module Name: src
> > > Committed By:riastradh
> > > Date:Fri Oct 19 02:07:23 UTC 2012
> > >
>
,
> ".gnupg/pubring.gpg");
> }
What happens if $HOME is undefined ?
What happens if $HOME is very long ?
Why did someone use ?: not if ?
David
--
David Laight: da...@l8s.co.uk
N
> buffer further down the call tree. If the buffer is truncated, the
> correct default public keyring will not be found. If the default
> public keyring is not found, the verification will fail.
Silent truncation seems a bad thing to do in security code.
David
--
David Laight: da...@l8s.co.uk
rwrite
the end of the malloced area.
You need to move the realloc() below the cond_depth+++
I'm then not sure that 128+n*32 is a sane sequence.
Possibly just n*32 - since I suspect 32 is plenty for most makefiles.
David
--
David Laight: da...@l8s.co.uk
probably almost
always never chaned (well maybe execpt default error values).
Marking function parameters 'const' is also pointless, some compilers
verify that the definition in the headers file matches - which is
partitularly pointless since the called object code isn't required
to avoid modifying the argument memory/register (which might sometimes
be a useful optimistation).
David
--
David Laight: da...@l8s.co.uk
Mayve it should verify that the last entry has been written to.
David
--
David Laight: da...@l8s.co.uk
ffer size, and document the
> buffer size we use. This allows us to cat -B 1000 /proc//maps
> for example which cannot handle seeking.
Isn't that what dd is for ?
David
--
David Laight: da...@l8s.co.uk
remove the main() declaration.
This might break some scripts.
David
--
David Laight: da...@l8s.co.uk
o!
Is pkill() guaranteed to do an atomic traversal of the process list?
So it will kill something that keeps using fork() to change its pid.
David
--
David Laight: da...@l8s.co.uk
gt; How could you guarantee this short of moving this into the kernel
> so you could do all the comparisons while keeping all forks from happening?
You probably can't ...
Even in the kernel it might be 'interesting'.
David
--
David Laight: da...@l8s.co.uk
l.h sysctl.h
> src/usr.bin/mail: head.c
>
> Log Message:
> Fix misspelling: accommodate is a long enough word to have room for two 'c's
> and two 'm's.
Two of everything except the 'd' and 't' - and there isn't room for two
of either of those.
David
--
David Laight: da...@l8s.co.uk
ALLED to no, now that
> all users of comapt/defs.mk have been adjusted. This is the last
> commit related to fixing PR 47188.
Phew :-)
David
--
David Laight: da...@l8s.co.uk
ne that might be due for relegation is NMCA, all the
rest will be defined anyway.
David
--
David Laight: da...@l8s.co.uk
no one has any more,
and really aren't of enough historical interest to keep supporting.
David
--
David Laight: da...@l8s.co.uk
s that isn't ususally true.
David
--
David Laight: da...@l8s.co.uk
by MI code.
Not only that, sys/proc.h is a posix specified header.
It really wants splitting into 3:
1) the bits posix says it should have, needed by userspace.
2) the kernel public bits, used by normal drivers
3) the bits that are private to the process handling code & scheduler.
David
--
David Laight: da...@l8s.co.uk
this change is useful for besides re-adding a host dependancy
> on sed.
It would be better to use make conditionals to fix the strings.
David
--
David Laight: da...@l8s.co.uk
ts destructors, paging in code and data just to exit!
About the only environment wheere is matters is when programs are
run as shell builtins - and that will always be a small subset of
programs.
David
--
David Laight: da...@l8s.co.uk
On Sun, Dec 16, 2012 at 11:15:21AM +, David Laight wrote:
> Module Name: src
> Committed By: dsl
> Date: Sun Dec 16 11:15:21 UTC 2012
>
> Modified Files:
> src/libexec/ld.elf_so: rtld.c
>
> Log Message:
> You need to pass 0 (not -1) to lwp_park() if
takes one cycle per skipped instruction.
> branch to unlikely cases instead of conditionally skipping them.
Or more for 'ldm' and 'stm' - esp. I think on strongarm!
Conditional execution might also be a full pipeline stall, rather
than a branch which is likely (hopefully) to
On Mon, Dec 31, 2012 at 03:00:42AM +0100, Joerg Sonnenberger wrote:
> On Sun, Dec 30, 2012 at 08:16:59PM +0000, David Laight wrote:
> > Module Name:src
> > Committed By: dsl
> > Date: Sun Dec 30 20:16:59 UTC 2012
> >
> > Modified Fil
no than ENOMEM.
Certainly checking for multiply overflow would seem better than
checking for the product being zero.
Unfortunately that check tends to need a divide - although some
simple range checks will avoid that in most cases.
David
--
David Laight: da...@l8s.co.uk
.c
> >
> > Log Message:
> > No need to check both TARGET_64BIT and ix86_preferred_stack_boundary >= 64,
> > if the former is true the latter is also true.
>
> can this not be overriden with cli options? the old
> test seems safer.
No - for amd64 the ix86_pre
clude "sys/ioctl.h" for the prototype of ioctl(2) which fixes the build.
That code (and zfs) seems to be built with rather generous compiler options.
David
--
David Laight: da...@l8s.co.uk
preferred stack boundary
> >
> > Eh? TARGET_64BIT forces ix86_preferred_stack_boundary >= 64.
> > The fact that it also implies SSE support is relevant.
>
> Only as default value, not if explicitly overriden?
It won't let you override to a lower value.
David
--
David Laight: da...@l8s.co.uk
On Thu, Jan 03, 2013 at 04:42:55PM +, David Laight wrote:
> Module Name: src
> Committed By: dsl
> Date: Thu Jan 3 16:42:55 UTC 2013
>
> Modified Files:
> src/external/cddl/osnet/sys/sys: random.h
>
> Log Message:
> Use cprgn_fast() not rnd_extract
ent
and double-word copies!
David
--
David Laight: da...@l8s.co.uk
e)
rather than relying on make's 'hunt for a source file' suffix rules.
Doing that would make it much easier for some sub-architectures to
fall back on the .c version even when a .S one exists (but, for example,
uses unavailable instructions).
David
--
David Laight: da...@l8s.co.uk
char * into a long, return NULL
> and set errno to ENOTSUPP.
Could the 'char *' pointers be replaced with indexes into an array?
David
--
David Laight: da...@l8s.co.uk
On Fri, Jan 25, 2013 at 04:06:09PM -0700, Warner Losh wrote:
>
> On Jan 25, 2013, at 4:05 PM, David Laight wrote:
>
> >> Log Message:
> >> For platforms where we cannot fit a char * into a long, return NULL
> >> and set errno to ENOTSUPP.
> >
> &
result you have to do far more than loop
for EAGAIN - adding such a loop is a bogus fix.
Not much point writing an error is you'vejust failed to write to stderr!
David
--
David Laight: da...@l8s.co.uk
'fix' was right.)
David
--
David Laight: da...@l8s.co.uk
NE_ARCH:Mearm*)
The former is much easier for make to parse :-)
David
--
David Laight: da...@l8s.co.uk
the RHS of assignments
when they are on local or export lines.
eg:
(x="a b"; b=fubar; export y=$x; echo $y; (echo $b))
outputs a and fubar (on two lines).
David
--
David Laight: da...@l8s.co.uk
On Mon, Feb 25, 2013 at 08:43:51PM +, David Laight wrote:
> On Mon, Feb 25, 2013 at 06:54:13PM +, David Holland wrote:
> > On Mon, Feb 25, 2013 at 06:49:51PM +, Julio Merino wrote:
> > > Log Message:
> > > Cherry-pick upstream change d0daf9983f5a0
o timeout' case (try to) check that the elapsed time is
less than one tick?
David
--
David Laight: da...@l8s.co.uk
On Sat, Mar 09, 2013 at 12:19:08AM +0100, Martin Husemann wrote:
> On Fri, Mar 08, 2013 at 11:03:23PM +0000, David Laight wrote:
> > Should the 'no timeout' case (try to) check that the elapsed time is
> > less than one tick?
>
> I gave the (partly virtual) test
negative errno numbers, not -1 which is interpreted
> as EPERM.
Hmmm... NetBSD usually uses +ve errno values.
David
--
David Laight: da...@l8s.co.uk
block (it does mutex_enter()),
> so we need to retry if curlwp took a context switch during the call.
I didn't think mutex_enter() blocked - isn't it a spinlock.
Which means that if things are going wrong they can go wrong
even if the mutex is available immediately.
David
--
David Laight: da...@l8s.co.uk
leep.
I know the uncontended path is now common to both (and fast) - unlike
lockmgr which took a couple of 100 instructions to decide what to do.
David
--
David Laight: da...@l8s.co.uk
so that the errno value can be returned in the same field.
They all tend to lead to coding errors!
David
--
David Laight: da...@l8s.co.uk
t say 1024 (dunno really).
Similarly for fds, maybe 256 and 4096.
At the moment there are nice local-user DoS because of the hard limit
values.
David
--
David Laight: da...@l8s.co.uk
ostic betin set - by pthread__init().
I think pthread_init() needs to be looking at some value that
libc set before calling main.
David
--
David Laight: da...@l8s.co.uk
list of buildable items.
That will look at the nodes far less often than the hack I had to do
to get .ORDER working.
David
--
David Laight: da...@l8s.co.uk
c)
and really doesn't want locale-specific alpha characters be valid.
(Otherwise scripts become non-portable.)
OTOH (unsigned)(ch - '0') <= 9 is probably the fastest isdigit()
on any modern cpu.
David
--
David Laight: da...@l8s.co.uk
to retire the BCS code on the
> > > next libc major bump.
> >
> > is it safe to make tools use host's strtoul?
>
> I think so. Tools shouldn't be using locales.
Do we try to force the C local at the top of build.sh?
Would probably be a good idea for anywhere where the script
looks at program output.
David
--
David Laight: da...@l8s.co.uk
4bit one that uses 32bit pointers?
A little unusual.
Doesn't that require a COMPAT_64 layer in the kernel?
Since user pointers will be 64bit and kernel ones only 32.
David
--
David Laight: da...@l8s.co.uk
INE_ARCH:Mearm*} != ""
Any reason?
Make finds it much harder to parse empty().
David
--
David Laight: da...@l8s.co.uk
s called twice, once with iocookie == NULL in order
to find out how big everything will be, then again with iocookie != NULL
to actually do the writes.
So changing the size of the register area on the second pass is going to
lead to corrupt core files.
David
--
David Laight: da...@l8s.co.uk
mine the label secton that can't be written to).
If you put a disk where 'c' is raw, and 'd' is u user partition into an x86
system the 'd' partition could be made available as wd0c!
David
--
David Laight: da...@l8s.co.uk
msr(MSR_TSC) != 0);
The:
+extern int cpu_msr_tsc;
looks bogus - debug ??
David
--
David Laight: da...@l8s.co.uk
olaris nor any SVR4 (or unixware) used FQDN.
A SUSE9 system I have at work (PII-450 and still used as an xterm)
gets it IP from DNS and hostname from a DHCP reverse lookup and
only uses the shortname.
David
--
David Laight: da...@l8s.co.uk
membered how snprintf works then. I didn't observe it in the wild
> but some dtrace code uses long vmem names ("dtrace_aggid_%d").
You might have remembered how the microsoft version of snprintf()
(doesn't) work!
David
--
David Laight: da...@l8s.co.uk
Fix compilation' would be better.
David
--
David Laight: da...@l8s.co.uk
gt; Add __le16, __le32, and __le64 to .
>
> As far as I'm aware, these aliases are correct, and the type doesn't
> actually have any effect on the byte order of reads and writes.
Correct - they are for the static analysis tools.
David
--
David Laight: da...@l8s.co.uk
age:
> Kludge around max as a local variable in intel_panel.c.
Possibly adding a #define for max/min before/after including the
netbsd header that defines these as a function/define would solve
this in the wrapper headers?
David
--
David Laight: da...@l8s.co.uk
es to use concatenated strings.
As well as avoiding the above problem it makes them much, much, much
easier to read.
I remember changing some of the source files to split the strings.
David
--
David Laight: da...@l8s.co.uk
ing 'netperf' which seems to accept absolutely any
crap on the command line, exceptionally difficult to determine what
it did.
David
--
David Laight: da...@l8s.co.uk
ollers.
Does NetBSD have a driver for them?
I presume it would have to be USB2 mode??
I've access to an adapter.
David
--
David Laight: da...@l8s.co.uk
rwlock their own cache-line. Also give a separate
> cache-line for the rwlock's reader counter.
You'd be much better off ordering the fields to avoid ping-pong
of data.
IIRC one of the new (big) ppc has 256 byte cache lines.
Cache line aligning data is then getting silly.
David
--
David Laight: da...@l8s.co.uk
bitches.
In that function isn't inlined then the generated code will be
extremely horrid.
The 'always inline' was there for a reason.
David
--
David Laight: da...@l8s.co.uk
On Sat, Oct 19, 2013 at 01:19:03AM +, John Nemeth wrote:
> Module Name: src
> Committed By: jnemeth
> Date: Sat Oct 19 01:19:03 UTC 2013
>
> Modified Files:
> src/sbin/gpt: gpt.8
>
> Log Message:
> type fix: accommodate. -> accomodate.
Why? The correct spelling has two 'm's
At
on above .code16 so that it produces an R_386_32 instead
> of an R_386_16 relocation, which is truncated to fit. XXX: untested.
Broken - it would have to be moved to before the call to prot_to_real.
David
--
David Laight: da...@l8s.co.uk
o {\
> (cpi)->intreg_4m->pi_set = PINTR_SINTRLEV(lvl); \
> (void)(cpi)->intreg_4m->pi_pend;\
> } while (0)
If pi_pend isn't volatile you need something like:
*(volatile * typeof ((cpi)->intreg_4m->pi_pend))&(cpi)->intreg_4m->pi_pend
You might choose to know the type :-)
David
--
David Laight: da...@l8s.co.uk
On Sun, Oct 20, 2013 at 07:49:55PM +0100, David Laight wrote:
> On Sat, Oct 19, 2013 at 08:16:16PM -0400, Christos Zoulas wrote:
> > Module Name:src
> > Committed By: christos
> > Date: Sun Oct 20 00:16:16 UTC 2013
> >
> > Modified
tes
> with character interpretation off as per SUSV2.
Does not expanding tabs actually make sense?
At a guess it would lead to tab characters being send to the terminal
and the display getting f**cked up.
I also suspect it will break applications that expect curses to
expand tabs.
David
--
David Laight: da...@l8s.co.uk
can cause serious problems because low priority threads can
quite reasonably hold sleep locks (that are contended by other similar
processes) for very long periods.
David
--
David Laight: da...@l8s.co.uk
quot;QF9700" USB 1.1 to 10Mbit/s Ethernet adapter.
> The multicast filter doesn't appear to work, so put the chip in promisc
> mode so that IPv6 NDP works.
Does it have a 'promiscuous for multicasts' option?
Possibly just setting all 1's in the filter will work.
David
--
David Laight: da...@l8s.co.uk
items on 32bit systems?
On i386 (and other le systems) this can be fixed be ensuring there
are 4 zero bytes following the data.
In sparc (and other be systems) it requires renaming of the symbol itself.
I presume such a rename was done when time_t was extended to 64bits?
David
--
David Laight: da...@l8s.co.uk
ove the vector
table to a safe address (the 286 ice didn't like it).
David
--
David Laight: da...@l8s.co.uk
ry"
constraint for the buffer areas either side of the actual copy.
David
--
David Laight: da...@l8s.co.uk
asm volatile ( :: "m" ( *(struct {char x[len]; } *)(ptr)) )
#define memcpy(d, s, len) \
if (__builtin_constant(len) && (len) == 8) { \
reference_memory(s, 8); \
*(uint64_t *)(d) = *(uint64_t *)(s); \
reference_memory(d, 8); \
} else \
__memcpy(d, s, len) \
But my gnu asm is a bit weak!
David
--
David Laight: da...@l8s.co.uk
manual unloads of drivers can happen when the
device is open - and that is also bad.
A manual unload probably isn't going to race with open or close though.
Disallowing unload completely would be a pain when developing drivers.
David
--
David Laight: da...@l8s.co.uk
ng TS.
Although I'm not 100% sure the TS is needed on a single cpu system.
It isn't there on amd64, but really it does no harm.
Without it one process will have an indererminate fp state.
I can't say I've actuallly looked a xen.
Presumably we somewhere compile the hypervisor as part of a kernel?
David
--
David Laight: da...@l8s.co.uk
is SMP capable. NetBSD dom0 currently aren't, but
> hopefully it is in the works.
Probably best to make fpuinit() xen-save then.
David
--
David Laight: da...@l8s.co.uk
On Sun, Feb 02, 2014 at 01:27:45PM +0100, Manuel Bouyer wrote:
> On Sat, Feb 01, 2014 at 10:49:44PM +0000, David Laight wrote:
> > On Sat, Feb 01, 2014 at 08:07:07PM +, Manuel Bouyer wrote:
> > > Module Name: src
> > > Committed By: bouyer
> > > D
On Sun, Feb 02, 2014 at 08:41:29PM +0100, Manuel Bouyer wrote:
> On Sun, Feb 02, 2014 at 06:53:55PM +0000, David Laight wrote:
> > Something needs to set the TS (task switched) flag when a new cpu
> > is added. Both amd64 and i386 'bare metal' have direct calls to
l_to_host (signal));
+ ptrace (request, pid, (PTRACE_TYPE_ARG3)1, sig);
You probably want to kill the comment from the last merge.
David
--
David Laight: da...@l8s.co.uk
On Wed, Feb 05, 2014 at 11:09:05PM +0100, Joerg Sonnenberger wrote:
> On Wed, Feb 05, 2014 at 06:52:22PM +0000, David Laight wrote:
> > Module Name:src
> > Committed By: dsl
> > Date: Wed Feb 5 18:52:22 UTC 2014
> >
> > Modified Files
reading to code properly):
no-sse => no-sse2 => no-sse3 => no-ssse3 & no-sse4a => no-sse4_1
=> no-sse4_2 => no-avx => no-fma => no-fma4
You either need all of them, and the one that the next sub-version
of the cpu/compiler add, or you assume that the first ones imply the
latter - at least to some degree.
David
--
David Laight: da...@l8s.co.uk
it to a memory location.
David
--
David Laight: da...@l8s.co.uk
they really shouldn't need.
Separation in the kernel is even worse.
David
--
David Laight: da...@l8s.co.uk
On Sun, Feb 23, 2014 at 02:48:19PM -0800, John Nemeth wrote:
> On Feb 23, 10:35pm, "David Laight" wrote:
> }
> } Module Name:src
> } Committed By: dsl
> } Date: Sun Feb 23 22:35:28 UTC 2014
> }
> } Modified Files:
> } src/sys
; >VM_MAXUSER_ADDRESS.
>
> Do these need to use the old style #define constants
> instead of the new style (since 2005) CTL_CREATE idiom?
I certainly thought that no new 'fixed number' sysctls were supposed to
be added.
David
--
David Laight: da...@l8s.co.uk
On Sat, Mar 01, 2014 at 08:31:42AM +0200, Alan Barrett wrote:
> On Thu, 27 Feb 2014, David Laight wrote:
> >Modified Files:
> > src/sys/kern: kern_sysctl.c
>
> >+case CTLTYPE_INT:
> >+/* Allow for 64bit read of 32bit value */
> >+
buffer dynamically using the size given by an initial
> sysctl() call with oldp = NULL.
Code that does that for a numeric value will be quite happy with
either a 32bit of 64bit result.
David
--
David Laight: da...@l8s.co.uk
ory address to xrstor.
Gah ... :-(
FWIW I managed to get gcc 4.8 to optimise some fp loops to use the ymm
registers - the xsave/xrstor code seemed to worn an amd64.
But I don't have an i386 install on a new enough system.
David
--
David Laight: da...@l8s.co.uk
r -d '"') && \
${HOST_CC} -E -I${SRCDIR} - | grep '@@@' | tr -d '" @') && \
${TOOL_SED} -e "s/@VERSION@/$$V/" < ${.CURDIR}/sqlite3.pc.in \
> ${.TARGET})
That would be safer.
Although you might want to remove the target file on failure.
Or maybe change the last line to:
> ${.TARGET}.tmp && mv ${.TARGET}.tmp ${.TARGET}
David
--
David Laight: da...@l8s.co.uk
is n = 0 or whether this
> API tries to be fancy and it is n = 1. 1U << n is just straigtforward.
Or number from the other end...
Indeed, I have to go away and find the definitions and then realise
that they are just longhand!
I don't normally compare bit masking against zero, just:
if (var & BIT)
or
if (!(var & BIT))
to me they read better that way.
David
--
David Laight: da...@l8s.co.uk
a time_t type time.
The value is actually an interval in microseconds and bounded to less than
a second.
The code is doing:
long x = arc4random() % (1 * 100);
That really shouldn't generate a compiler warning.
David
--
David Laight: da...@l8s.co.uk
On Tue, Mar 18, 2014 at 08:21:19PM +, Christos Zoulas wrote:
> In article <20140318201420.go20...@snowdrop.l8s.co.uk>,
> David Laight wrote:
> >On Tue, Mar 18, 2014 at 03:30:09PM -0400, Christos Zoulas wrote:
> >> Module Name: src
> >> C
- so need not emit a warning.
For instance with:
for (int i = 0; i < sizeof foo; i++)
it can't matter that the comparison is unsigned because 'i' can be
assumed to be non-negative (even if 'sizeof foo' is greater than MAXINT).
David
--
David Laight: da...@l8s.co.uk
implicitly cast to uint32_t next.
Gah - I hadn't spotted that xtos had added a cast as well.
There is another cast that ought to be pointless two lines later.
David
--
David Laight: da...@l8s.co.uk
201 - 300 of 350 matches
Mail list logo
