; len = MBLKL(mc);// returns 0 ???
>
>
> -sf
> --
> This message posted from opensolaris.org
> ___
> opensolaris-code mailing list
> opensolaris-code@opensolaris.org
> http://mail.opensolaris.org/mailman/listinfo/opensolaris-code
such that minors that are used for the
filesystem nodes can be easily excluded from the arena.
Paul
--
===
Paul Durrant
http://www.linkedin.com/in/pdurrant
===
___
opensolaris-code ma
- hence my comment
about stashing the context pointer there.
Paul
--
=======
Paul Durrant
http://www.linkedin.com/in/pdurrant
===
___
opensolaris-code mailing list
opensolaris-
lookup is best left in the hands of the
driver (since it 'knows' its dev_t space and possibly its usage profile).
Having a context pointer in the vnode (or snode) OTOH is simple and
cheap, but does require an interface change.
Paul
--
=======
Paul Durrant
h
in the struct vnode?
Paul
--
=======
Paul Durrant
http://www.linkedin.com/in/pdurrant
===
___
opensolaris-code mailing list
opensolaris-code@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/opensolaris-code
n unload.
Paul
--
=======
Paul Durrant
http://www.linkedin.com/in/pdurrant
===
___
opensolaris-code mailing list
opensolaris-code@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/opensolaris-code
--
===
Paul Durrant
http://www.linkedin.com/in/pdurrant
===
___
opensolaris-code mailing list
opensolaris-code@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/opensolaris-code
yield the CPU though. I'd suggest using drv_usecwait() for
anything up to the millisecond level and then perhaps cv_timedwait() (or
if you don't care about using published interfaces, you could use
realtime_timout() directly).
Paul
--
===
Paul Du
ernel or user space? timeout is a kernel function (9f) and timer_create
is a user function (3rt).
Paul
--
===
Paul Durrant
http://www.linkedin.com/in/pdurrant
===
___
opensolaris-code ma
ou want to
attach a callback to be kicked off by a cyclic then I guess you'll need
to come up with your own mechanism.
Paul
--
=======
Paul Durrant
http://www.linkedin.com/in/pdurrant
===
___
James Carlson wrote:
> Paul Durrant writes:
>> James Carlson wrote:
>>> I had this problem a while ago, and it turned out that I was using a
>>> stale copy of the 'on-closed' tarball. Could that be your problem as
>>> well?
>>>
>> That&
James Carlson wrote:
>
> I had this problem a while ago, and it turned out that I was using a
> stale copy of the 'on-closed' tarball. Could that be your problem as
> well?
>
That's quite likely. I'm build a recent ON but the most recent non-DEBUG
on-closed I can find is the snv_105 set, since
Hi,
I'm running a nightly build on a recent (a couple of days ago) pull
from onnv clone and I'm hitting the following error:
Failed to create generic kernel archive:208450 blocks
cpiotranslate: kernel/brand/amd64: no packaging info
cpiotranslate: kernel/brand: no packaging info
It
Christian Kaiser wrote:
>
> I wonder if there are cache allocation operations available for DMA memory?
>
> What I am searching would be something like kmem_cache_*(9F) but for DMA
> memory and not for any kernel memory.
>
In a word, no. I did put together a proposal for DMA cache allocators
Mahesh wrote:
>
> I am in the process of implementing a GLD v3 driver for open solaris
> and i would like to know the some info on LRO support in the Solaris
> stack.Does it provide some hooks for Software based/Hardware based
> LRO ?? .
No. There is no support in the stack.
> I read in some of
Christian Kaiser wrote:
>> Presumably, also, if you go look in /etc/devaliases you only see one
>> entry for your driver?
>
> I don't have such a file here.
> If you meant /etc/devlink.tab... yes.
>
No. Sorry, I mean /etc/driver_aliases
>> Is there a call to detach(9e) between the two attach(9e
Christian Kaiser wrote:
>
> But... unfortunately I still see the two attach calls for instance #0.
> Any idea how to debug this?
>
After it is rebuilt you definitely only have one entry in
/etc/path_to_inst? Presumably, also, if you go look in /etc/devaliases
you only see one entry for your d
Christian Kaiser wrote:
>
> But in the syslog I see two calls to my attach function for the same
> instance #0. What other place may store the information about the second
> (although non-existent) instance #0 then?
>
Did you rebuild your boot archive? /etc/path_to_inst is included in the
boo
Christian Kaiser wrote:
>
> Thanks Paul! That helped a lot!
> But there is still one minor thing confusing me. Why is ddi_dev_nregs()
> returning nregs=6 and not nregs=5?
>
The brutally simple answer is that the implementation of that function
simply takes the length of the 'reg' property and
Christian Kaiser wrote:
> OK... I dont't understand this at all. I did a 32-bit read on all BARs.
> Maybe you can explain me this?
>
> 0x10 0xc830
> 0x12 0xbeefc830
> 0x14 0x0
> 0x16 0x0
> 0x18 0xd800
> 0x20 0xd000
> 0x22 0xbeefd000
> 0x24 0xcc00
> 0x26 0xcc00
>
Those are 32-bi
sreenatha wrote:
> I am doing some applications and wanted to know about:
> How do we get the information like IO mapped IO(which port address is
> assigned to which device)
> and memory mapped IO(what memory range is assigned to which device) in
> solaris
> as is given by /proc/ioports and
wan_jm wrote:
> I want to find the source code for /usr/sbin/df , but I don't know to search,
> use which key word and what project is it in?
> could you please help me?
Try
http://src.opensolaris.org/source/search?q=&defs=&refs=&path=df&hist=&project=%2Fonnv
__
Jason King wrote:
> I'm curious to know if anyone feels if there might be utility in
> providing macros in a header file for defining and manipulating
> bitmasks (analogous to __BIT(x), __BITS(a,b), __SHIFTIN, __SHIFTOUT,
> etc on some of the BSDs)? In some of the driver work I've done, I've
> fou
mes in my
recent memory, but it brings with it a lot of challenges and usually
kills system performance.
Paul
--
Paul Durrant
http://www.linkedin.com/in/pdurrant
___
opensolaris-code mailing list
opensolaris-code@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/opensolaris-code
xulari wrote:
> thanks for advice!
> anyway,is there a way to load the driver once ,and system will not
> detach the diver?
>
Yes, write your driver to return failure from it's detach(9e) routine:
it's then no longer possible for the system to detach and unload your
driver.
(If you ev
xulari wrote:
> I wrote a driver xxx, I want it to be loaded when i start the computer
> so i put xxx into /kernel/drv/sparcv9/xxx
> but after i reboot the solaris10, i check the /var/adm/messages
> i found just after driver "attach", it "detach" ...
> then i
> "add_drv xxx" failed , instructions "
Joachim Worringen wrote:
> As it isn't mentioned by the man page, and "Solaris Internal" makes only
> an implicit statement (first sentence of 11.2.3.7, p.547): is the memory
> returned by kmem_alloc() always physically contiguous?
>
No.
Paul
___
; layers thin and simple.
>
Strangely enough our s/w stack has to run on Linux, Windows and
Solaris too and the mismatching kernel APIs and threading models are a
pain ;-/
Paul
--
Paul Durrant
http://www.linkedin.com/in/pdurrant
___
opens
ll work for you. As I said though, IMO the
thread interface should be made public. In fact, it would be
interesting to look at what interfaces are commonly used by drivers in
Linux or Windows and make sure Solaris has similar functionality; the
easier porting drivers is, the more people will
clics
that could be arrived it. However, you *can* create a thread;
thread_create() will do it for you (I think it's in kthread.c). I
don't believe it's in the DDI, but IMO it should be as it can be
highly useful.
Paul
--
Paul Durrant
http://www.linkedin.com/in/pdurrant
_
uired to do cv_signal.
Ok. That makes perfect sense :-)
Paul
--
Paul Durrant
http://www.linkedin.com/in/pdurrant
___
opensolaris-code mailing list
opensolaris-code@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/opensolaris-code
not found
Have you written an mdb module too? ::load is for loading mdb modules
(see ::help load).
Paul
--
Paul Durrant
http://www.linkedin.com/in/pdurrant
___
opensolaris-code mailing list
opensolaris-code@opensolaris.org
http://mail.opensolaris.or
On 29/05/07, [EMAIL PROTECTED] <[EMAIL PROTECTED]> wrote:
Paul Durrant wrote:
> On 28/05/07, Garrett D'Amore <[EMAIL PROTECTED]> wrote:
>> You should not attempt to acquire a lock which will be held by other
>> functions calling cv_wait, or its brethren. (The
ot aware of this restriction, nor can I find it documented.
This suggests that one cannot call cv_signal() from within an
interrupt handler. Is this true?
Paul
--
Paul Durrant
http://www.linkedin.com/in/pdurrant
___
opensolaris-code mailing list
opens
On 21/05/07, Thomas De Schampheleire <[EMAIL PROTECTED]> wrote:
The exact location (/proc or elsewhere) is no point for me, but I do
find this mechanism an extremely useful feature.
You could try running mdb -wk and then patching globals in your module
from there.
Paul
--
Paul D
to see if you get to one of the core DDI guys.
Paul
--
Paul Durrant
http://www.linkedin.com/in/pdurrant
___
opensolaris-code mailing list
opensolaris-code@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/opensolaris-code
.
You'll need the -N option in the linker for the first item too.
Remember that globals and non-static functions are both examples of
exported symbols.
Paul
--
Paul Durrant
http://www.linkedin.com/in/pdurrant
___
opensolaris-code mailing list
;s probably a
private function which you should consider unstable.
Secondly, I am still stuck with the question about how my module will be
loaded. It should be loaded by default when booting, is there a way to
achieve that?
What type of module is it? A driver? If so, I think ddi-forceattach
_fini()?
Paul
--
Paul Durrant
http://www.linkedin.com/in/pdurrant
___
opensolaris-code mailing list
opensolaris-code@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/opensolaris-code
On 4/30/07, Frank Hofmann <[EMAIL PROTECTED]> wrote:
Hmm - ddi_enter_critical / ddi_exit_critical ?
That certainly sounds like it would do the job. Thanks :-)
Paul
--
Paul Durrant
http://www.linkedin.com/in/pdurrant
___
opensolaris-code m
time) what
mechanism could I use - within the Solaris DDI - to make sure a thread
accessing the device was not pre-empted thus causing such a timeout?
Outside of the DDI I have a few options it seems;
kpreempt_disable/enable or splhigh/splx perhaps. Within the DDI I've
not yet found anything.
om within a task handler, so you
should be able to use taskqs or timeout to achieve what you want but
using both is unnecessary.
Paul
--
Paul Durrant
http://www.linkedin.com/in/pdurrant
___
opensolaris-code mailing list
opensolaris-code@opensolar
ave certain
advantages/disadvantages?
You could also look at the interface in cyclic.h. The only problem
with this is that it's not a public interface and so your build env.
would be tied to ON.
Paul
--
Paul Durrant
http://www.linkedin
to avoid an LD_PRELOAD. I've had to pull such
tricks in the past when writing my own preloads.
Paul
--
Paul Durrant
http://www.linkedin.com/in/pdurrant
___
opensolaris-code mailing list
opensolaris-code@opensolaris.org
http://mail.opensolaris.org/
t;movl %%eax, 0(%%ebx)\n"
"popl %%ebx\n"
: /*wr*/
: /*rd*/"g" (block)
: /*trash*/ "eax", "edi", "ecx", "edx", "memory", "cc"
);
}
Does anyone know where i am wrong?
Thank you!
Regard
ns I'm using the "tip".
Yes, until you do:
hg update onnv_59
at which point you'll be using onnv_59.
Paul
--
Paul Durrant
http://www.linkedin.com/in/pdurrant
___
opensolaris-code mailing list
opensolaris-code@opensolaris.org
t from opensolaris.org
--
Paul Durrant
http://www.linkedin.com/in/pdurrant
___
opensolaris-code mailing list
opensolaris-code@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/opensolaris-code
n't help you much. How did you set up your
environment. If you run /opt/onbld/bin/bldenv -d then you
should get a DEBUG environment i.e. -DDEBUG will be defined on the
compiler command line. Hence any code bracketed by #ifdef DEBUG/#endif
will be compiled in.
Paul
--
Paul Durrant
http
c' interface - see usr/src/uts/common/sys/mac.h
Paul
--
Paul Durrant
http://www.linkedin.com/in/pdurrant
___
opensolaris-code mailing list
opensolaris-code@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/opensolaris-code
up(). There's a nexus upcall which uses the reg and
assigned-addresses props to do figure out the mapping. If you can't
sacrifice an existing BAR and want to add a BAR, assuming you can find
a spare bit of physical address space then you should be able to tag
on an entry to the reg and assi
duling.
What would be useful to me is a means of kicking a soft interrupt of
on a given CPU (I assume they run on the same CPU as the caller of
'trigger' at the moment) so I can get work done where I know the data
should be in cache.
Paul
On 22 May 2006, at 19:26, Jack Schwartz wrote:
This was fixed for X86 as well, but I believe this was done the day
S10U1 released. The fix went into S10U2. X86 S10U1 is missing this
definition.
X86 KU patch 118844-27.
Bug ID: 6356171
RTI: 305236
Thanks all.
Paul
--
Paul Durrant
dditypes.h so my
question is, is the def. truly missing in s10 01/06 or am I missing a
patch?
Paul
--
Paul Durrant
http://www.linkedin.com/in/pdurrant
___
opensolaris-code mailing list
opensolaris-code@opensolaris.org
http://mail.opensolaris.org/mailman
he def. of pthread_mutex_init() in open.c and you'll
notice the #ifdef _WINDOWS]
--
Paul Durrant
PGP.sig
Description: This is a digitally signed message part
___
opensolaris-code mailing list
opensolaris-code@opensolaris.org
https://opensolaris.
the idea of some sort of message
saying what a library does is quite a cool one.
Paul
--
Paul Durrant
PGP.sig
Description: This is a digitally signed message part
___
opensolaris-code mailing list
opensolaris-code@opensolaris.org
https://ope
port Linux application and want to
implement the same workaround.
Can you not use SO_LINGER?
--
Paul Durrant
PGP.sig
Description: This is a digitally signed message part
___
opensolaris-code mailing list
opensolaris-code@opensolaris.org
56 matches
Mail list logo
