Memory corruption bug in fdt_generic_util.c

[Bill-Paul]

Memory corruption bug in fdt_generic_util.c

[alistair23]

Hello Bill-Paul,

First of all I'm glad you decided to use QEMU I'm sorry you had difficulties building and running it. Internally we build, test and develop on Linux machines. Which is why there is little to no consideration on running on BSD or Windows hosts. Recently we have started adding some support for Windows hosts, although this is limited at the moment.

We know the code base is starting to age on targets and devices that we don't use and we are working on fixing this. Hopefully we will have a newer QEMU release later in the year.

1. I completely agree with you. I inherited this inability to build other platforms and just haven't had the time to spend on fixing it. We are working on improving the QEMU code though.
2. I'm sorry about that. Hopefully when we update our tree based on a newer mainline version this will be fixed.
3. Do you want to write a patch for this?
4. I can look into this and see what I can do to fix the problem.

Wow! Good catch. I can't believe we haven't seen this before. Thanks for pointing this out. I will write-up a fix.

Thanks for the feedback, we are always happy to hear it. We are trying hard to improve QEMU and any input is appreciated.

Thanks,
Alistair

[Bill-Paul]

Regarding the lseek64()/lseek() thing, sadly I don't really have a patch, because I'm not exactly certain what the correct fix would be. For now I changed the source to just use lseek() so that it would build on FreeBSD. I was told the correct way to address this is:

#include <unistd.h>
...
#if (_POSIX_V6_ILP32_OFF32 > 0) || (_POSIX_V7_ILP32_OFF32 > 0)
...lseek64(...
#else
...lseek(...
#endif

But I'm not 100% sure of this.

It's possible that for the other machine types the number of inputs never exceeds 15, which means it could only be an issue with the MPSoC simulation. You can see it by adding a printf() in qemu_irq_shared_handler() to display the value of 'n' each time it's called. With the zcu102 machine description it definitely shows values larger than 15. I built the code on an Ubuntu machine too just to make sure.

Anyway, with the fixes in place, I was able to run the simulator on my FreeBSD host and boot the pre-compiled ZCU102 U-Boot and Linux images, so I'm good for now. I have been using QEMU to for VxWorks development and debugging for quite some time and it's always come in very handy, so I was very interested in using it for the Xilinx platforms too.

-Bill

[alistair23]

Great! I'm glad it works for you after those issues.
I'll path the memory issue early next week and then I'll try to look into the lseek problem as well.
Thanks,
Alistair