Re: [ECOS] ARM - out of MBUFS - allowing FIQ's in splimp

Re: [ECOS] ARM - out of MBUFS - allowing FIQ's in splimp

[AshCan]

I brought over the latest tcpip code, but did not see any driver code.  Is this right?  

Also, I am having problems making the kernel based on the old version drivers and latest anon-cvs tcpip.  

I brought the tcpip into a new directory called current parallel with v1_0b1 and changed the correspnding line in my ecos.ecc, but the kernel build fails on file drivers/eth/edb7xxx/v1_0b1/src/if_eth7xxx.c.  All kinds of problems with include files, invalid tokens, .... My other ecos.ecc build items point to the v1_0b1 driver directories. 

I will eventually figure this out, but does anyone know what is wrong?

thanks,

Dan Ash

AshCan@aol.com wrote:
> 
> Is it possible this panic condition has been caused by my allowing DAC
FIQ's?
> 
> Are there any other possible reasons for this panic condition?

Which sources are you using? I believe the sources in the net EPK on the
web are susceptible to this, but that this has since be fixed in anonymous
CVS. So if you're not already using anon CVS, try that and see if it makes
a difference.

Re: [ECOS] ARM - out of MBUFS - allowing FIQ's in splimp

[Jonathan Larmour]

AshCan@aol.com wrote:
> 
> I brought over the latest tcpip code, but did not see any driver code.  Is this right?

The driver code has now moved into the main eCos sources. You can only use
the latest tcp/ip code in anonymous CVS with the other eCos code also
available from anonymous CVS.

Jifl
-- 
Red Hat, 35 Cambridge Place, Cambridge, UK. CB2 1NS  Tel: +44 (1223) 728762
"Plan to be spontaneous tomorrow."  ||  These opinions are all my own fault

Re: [ECOS] ARM - out of MBUFS - allowing FIQ's in splimp

[Jonathan Larmour]

AshCan@aol.com wrote:
> 
> Is it possible this panic condition has been caused by my allowing DAC FIQ's?
> 
> Are there any other possible reasons for this panic condition?

Which sources are you using? I believe the sources in the net EPK on the
web are susceptible to this, but that this has since be fixed in anonymous
CVS. So if you're not already using anon CVS, try that and see if it makes
a difference.

Jifl
-- 
Red Hat, 35 Cambridge Place, Cambridge, UK. CB2 1NS  Tel: +44 (1223) 728762
"Plan to be spontaneous tomorrow."  ||  These opinions are all my own fault

[ECOS] ARM - out of MBUFS - allowing FIQ's in splimp

[AshCan]

I am using the EDB7212 board and TCP/IP.  I am playing audio and since the EP7212 serial ports only have 8 deep FIFOs, they have to be fed often (half empty FIQ every 100us with a 44.1KHz sampling rate).

Because of this, I was missing DAC FIQ's, thus samples, when doing TCP/IP transmit.  

So I went into HAL_DISABLE_INTERRUPT and made it only disable IRQ's, still allowing FIQ's.  This corrected my DAC problem and no more samples were missing.  The effect was no more glitches in the audio playback.

Now, after hours of a single socket connection, I get "out of MBUFS" panics.  This traces back to the eth_drv_recv() function.  

Is it possible this panic condition has been caused by my allowing DAC FIQ's?

Are there any other possible reasons for this panic condition?

My DAC ISR only uses ARM registers r8-r14.

thanks for any help,

Dan Ash
SonicBox