What is "asy1 : Silo Overflow"?
Ken Tan
Anyone know why I get the "asy1 : Silo Overflow" msg when I do PPP
connections on Solaris 2.5.1 x86? Port speed is 38400 with a 33,600
modem... anyone know how to fix this?
Much Appreciated..
Ken
Andrew Gabriel
In article <01bc7a48$01cbd1e0$3d4818d2@smurfie>,
"Ken Tan" <s...@post1.com> writes:
>Anyone know why I get the "asy1 : Silo Overflow" msg when I do PPP
>connections on Solaris 2.5.1 x86? Port speed is 38400 with a 33,600
>modem... anyone know how to fix this?
From the asy man page:
DIAGNOSTICS
asyn: silo overflow.
The hardware overrun occurred before the input
character could be serviced.
This means the driver isn't handling interrupts fast enough.
I suspect that either you don't have a buffered UART, or that
you are using Driver Update 7 (which fails to turn the buffer
on), or that you have a third party device driver installed
which is keeping interrupts disabled too long.
It might be useful to know what type of processor and clock
speed you are running, but Solaris on a 486 33Mhz has no trouble
keeping up with a 57600 port, so that's unlikely to be the problem.
--
Andrew Gabriel And...@cucumber.demon.co.uk
Consultant Software Engineer
Paul David Fox
Andrew Gabriel wrote:
>
> In article <01bc7a48$01cbd1e0$3d4818d2@smurfie>,
> "Ken Tan" <s...@post1.com> writes:
> >Anyone know why I get the "asy1 : Silo Overflow" msg when I do PPP
> >connections on Solaris 2.5.1 x86? Port speed is 38400 with a 33,600
> >modem... anyone know how to fix this?
>
> From the asy man page:
>
> DIAGNOSTICS
> asyn: silo overflow.
> The hardware overrun occurred before the input
> character could be serviced.
>
> This means the driver isn't handling interrupts fast enough.
> I suspect that either you don't have a buffered UART, or that
> you are using Driver Update 7 (which fails to turn the buffer
> on), or that you have a third party device driver installed
> which is keeping interrupts disabled too long.
None of these. I have a dual PPro with buffered UARTs and DU8
and no 'special' hardware. It happens occasionally to me as well.
Normally its not a problem.
But on my system (2.5/x86) the system will occasionally lock
something in the kernel and never unlock it: Symptom: the
poll() with timeout system call stops working.
I usually have to reboot when this happens and it happens only
when I'm using the modem - very sporadically, even when there is
no load on the system.
> It might be useful to know what type of processor and clock
> speed you are running, but Solaris on a 486 33Mhz has no trouble
> keeping up with a 57600 port, so that's unlikely to be the problem.
Yeah - but the people at Sun have never bothered to test it
further than 'Oh - its worked for 5 seconds, lets ship it'.
If they used it for PPP net access they wouldnt have the
shambles they currently have upto and including DU8.
(I have seen no evidence to suggest DU9 is any better in this area
so I'm waiting for 2.6).
--
______________________________________________________________________
| This signature is deceased. RIP. It is an ex-signature. |
| 1996-1997 - a dearly beloved friend. |
______________________________________________________________________
Joerg Schilling
In article <33A6352B...@lehman.com>,
Paul David Fox <pfoxS...@lehman.com> wrote:
>Andrew Gabriel wrote:
>>
>> In article <01bc7a48$01cbd1e0$3d4818d2@smurfie>,
>> "Ken Tan" <s...@post1.com> writes:
>> >Anyone know why I get the "asy1 : Silo Overflow" msg when I do PPP
>> >connections on Solaris 2.5.1 x86? Port speed is 38400 with a 33,600
>> >modem... anyone know how to fix this?
>>
>> From the asy man page:
>>
>> DIAGNOSTICS
>> asyn: silo overflow.
>> The hardware overrun occurred before the input
>> character could be serviced.
>>
>> This means the driver isn't handling interrupts fast enough.
>> I suspect that either you don't have a buffered UART, or that
>> you are using Driver Update 7 (which fails to turn the buffer
>> on), or that you have a third party device driver installed
>> which is keeping interrupts disabled too long.
>
>None of these. I have a dual PPro with buffered UARTs and DU8
>and no 'special' hardware. It happens occasionally to me as well.
>Normally its not a problem.
silo overflow simply means that the interrupt latency was too high
****at this moment**** , for what reason ever.
Joerg
--
EMail:jo...@schily.isdn.cs.tu-berlin.de (home) Jörg Schilling D-13353 Berlin
j...@cs.tu-berlin.de (uni) If you don't have iso-8859-1
j...@fokus.gmd.de (work) chars I am J"org Schilling
URL: http://www.fokus.gmd.de/usr/schilling ftp://ftp.fokus.gmd.de/pub/unix
Paul David Fox
Joerg Schilling wrote:
>
> silo overflow simply means that the interrupt latency was too high
> ****at this moment**** , for what reason ever.
>
Yes I know this; but there is a race-condition bug in the Solaris
kernel which bites randomly and frequently under non-load conditions
(non-load: high sustained RS-232 load but no other system activity).
Silo overflows occur when they probably shouldnt implying that
either there are some long execution paths inside the Solaris kernel
(I could understand that on a 486 but not a PPro unless they;ve
bloated the kernel over the years), or something is forgetting
to enable interrupts and only some other thing is causing them
to be enabled later on 'almost by accident'.
Anyone who's programmed RS-232 interrupts hopefully knows what I
mean.
