Linux SDK V1.3 problem

[matthewh]

I'm working with the Linux SDK V1.3 on a P2041 processor project and have hit an unexpected issue.  Using the P2041 system as a client, I mounted an NFSv3 filesystem from another server and started doing file reads and get a kernel panic 'BUG: scheduling while atomic: kworker/0:1/10/0x10000100' (note that the specified task running when the crash can differ, as I've seen 'kswapd0' and 'top' also).  I've gone back to a fairly generic SDK build (few if any modifications to kernel configuration, device tree, or drivers) and am still seeing this issue.  I've seen this issue using both the 1G and 10G MACs for server communication and have not seen any other kernel crashes using this ethernet connection before.

Has anyone else seen this type of crash and is there a known patch that might address it?

Thank you for helping with this,

MatthewH

[genuap]

So, you're using the P2041 as an NFS client (i.e. mounting an NFS share that exists on a different machine)?

I don't have access to a P2041, but I've tried this on a P4080 with no issue.I'm currently on SDK 2.3, but I'll try with SDK 1.3 to make sure it wasn't something fixed.

[matthewh]

Thanks for checking on this.  I've done some more exploring and am not sure the problem was in the NFS client.  However, I was doing high load tests on the NFS client code (and not much else) when the kernel crashes occurred.

Is there an expected release date for SDK 2.3?

[genuap · Accepted solution]

My mistake - SDK 1.3 is the latest. I keep going back and forth between the SDK's and older BSPs (before we had a unified SDK). The numbering is very different.

You can always get the latest off of the website:

http://www.freescale.com/webapp/sps/site/prod_summary.jsp?code=SDKLINUX

or off of the freescale git:

Freescale Public GIT