Re: [Qemu-devel] [PATCH v5 00/24] Fixing record/replay and adding reverse debugging

[dovgaluk]

Hi, Ciro!

I found several issues in your command lines.

Ciro Santilli писал 2018-08-08 02:13:
> OK, finally got some time to try it out, I'm using
> c42634d8e3428cfa60672c3ba89cabefc720cde9 from rr-180725.
>
> Replay works well as far as I can tell, so I moved to the reverse
> debugging:
>
> /home/ciro/bak/git/linux-kernel-module-cheat/out/x86_64/buildroot/build/host-qemu-custom.rr/x86_64-softmmu/qemu-system-x86_64
> \
> -M pc \
> -append 'root=/dev/sda nopat console_msg_format=syslog nokaslr
> norandmaps printk.devkmsg=on printk.time=y console=ttyS0 -
> lkmc_eval_base64="L3JhbmRfY2hlY2sub3V0Oy9wb3dlcm9mZi5vdXQ7"' \
> -kernel
> '/home/ciro/bak/git/linux-kernel-module-cheat/out/x86_64/buildroot/build/linux-custom.default/arch/x86/boot/bzImage'
> \
> -m '256M' \
> -monitor 'telnet::45454,server,nowait' \
> -nographic \
> -serial mon:stdio \
> -smp '1' \
> \
> -drive
> 'file=/home/ciro/bak/git/linux-kernel-module-cheat/out/x86_64/buildroot/images/rootfs.ext2.qcow2,format=qcow2,if=none,id=img-direct'

You'll probably need an overlay, it you want this file to be unchanged 
by VM.

Can you also provide this file for testing? I found only bzImage.

> \
> -drive driver=blkreplay,if=none,image=img-direct,id=img-blkreplay \
> -device ide-hd,drive=img-blkreplay \
> \
> -object filter-replay,id=replay,netdev=net0 \
> -device rtl8139,netdev=net0 \
> -netdev
> 'user,hostfwd=tcp::45455-:45455,hostfwd=tcp::45456-:22,id=net0' \
> \
> -icount
> 'shift=7,rr=record,rrfile=/home/ciro/bak/git/linux-kernel-module-cheat/out/x86_64/qemu/0/rrfile'

You need to specify rrsnapshot=<name> option for creating the initial VM 
snapshot.
This option creates snapshot at record and loads it at replay. GDB can 
also use this snapshot for reverse execution.

> \
>
> and replay with:
>
> -icount
> 'shift=7,rr=replay,rrfile=/home/ciro/bak/git/linux-kernel-module-cheat/out/x86_64/qemu/0/rrfile'
> \
> -gdb 'tcp::45457' \
> -S \
>
> Then, I do
>
> /home/ciro/bak/git/linux-kernel-module-cheat/out/x86_64/buildroot/host/usr/bin/x86_64-linux-gdb
> \
>  -q \
> -ex 'add-auto-load-safe-path
> /home/ciro/bak/git/linux-kernel-module-cheat' \
> -ex 'file vmlinux' \
> -ex 'target remote localhost:45457' \
> -ex 'break start_kernel' \
>   -ex continue \
> -ex 'lx-symbols ../kernel_module-1.0/' \
>
> Then in GDB:
>
> n
> n
> n
> n
> reverse-continue
>
> expecting it to return me to start_kernel, but instead it left me in
> the same place that I'm at.

Right, because there were no checkpoints. The initial one must be 
created at recording phase.



Pavel Dovgalyuk