Re: qemu-6.0.0: Assertion 'p_rcu_reader->depth != 0' failed

[Peter Maydell]

On Wed, 14 Jul 2021 at 01:44, Aileen Honess <aileen@brekersystems.com> wrote:
> I’m attempting to run multicore bare-metal tests on the qemu-6.0.0 release, 
> built from the downloaded tar file.  The tests themselves are generated, but 
> are a mix of C and ASM, compiled into an ELF. I’m migrating to the current 
> QEMU release from version 4.1.0 or so.
>
>
>
> My command line (with or without “-s -S”) is:
>
>
>
> %> /tools/gnu/qemu-6.0.0/bin/qemu-system-aarch64 \
>
>      -machine virt-6.0,secure=on,virtualization=on,gic-version=3 \
>
>      -s -S \
>
>     -m 1024M \
>
>     -cpu max \
>
>     -smp 4 \
>
>      -nographic \
>
>     -semihosting \
>
>     -bios pssTop.elf \
>
>     -device loader,file=pssTop.elf,addr=0 \
>
>     -device loader,addr=0x40100000,cpu-num=0 \
>
>      -device loader,addr=0x40100000,cpu-num=1 \
>
>      -device loader,addr=0x40100000,cpu-num=2 \
>
>      -device loader,addr=0x40100000,cpu-num=3;
>
>
>
> This will run and produce expected results (through the UART) for a bit.  The 
> amount of output varies from run to run.
>
> However, it will consistently die with this error, for which I find little 
> help:
>
>
>
> qemu-system-aarch64: /home/aileen/Downloads/qemu-6.0.0/include/qemu/rcu.h:93: 
> rcu_read_unlock: Assertion `p_rcu_reader->depth != 0' failed.

If QEMU asserts this is generally a bug. Could you provide the test
binary that reproduces this crash, please ?

Side note: why are you loading the 'pssTop.elf' both in -bios and
then again via the -device loader? I think the -bios option is
unnecessary here.

thanks
-- PMM