Obtaining trace from QEMU MBR disk image

Let $TRACE_DIR be a destination directory of CTF trace on host. Let /my_trace be the location of recorded trace on QEMU disk image (collected, e.g. via perf -m trace -o /my_trace)

The manual way

1. Find out the disk sector size and rootfs partition start (here, start=4096, sector size=512):

   user $ fdisk -l $PRTOS_PROJECT/_boot/ia32-generic-qemu/hd0.disk

   Disk _boot/ia32-generic-qemu/hd0.disk: 128 MiB, 134217728 bytes, 262144 sectors

   Units: sectors of 1 * 512 = 512 bytes

   Sector size (logical/physical): 512 bytes / 512 bytes

   I/O size (minimum/optimal): 512 bytes / 512 bytes

   Disklabel type: dos

   Disk identifier: 0x00000000

   Device Boot Start End Sectors Size Id Type

   _boot/ia32-generic-qemu/hd0.disk1 * 4096 262143 258048 126M 83 Linux

2. Mount the QEMU image:

   root # mount -o loop,offset=$((4096*512)) _boot/ia32-generic-qemu/hd0.disk /mnt

3. Copy the recorded trace directory to host destination directory:

   user $ cp /mnt/my_trace $TRACE_DIR

4. Proceed to Step 3: invoke convert.sh.

The automated way using rootfs_convert.sh

rootfs_convert.sh is an end-user script that does the above steps but uses parted instead (as it is more script-friendly), invokes convert.sh automatically and outputs a ready to view perfetto trace. Invoke

user $ cd $PRTOS_PROJECT/phoenix-rtos-hostutils/trace

user $ ./rootfs_convert.sh ../../_boot/ia32-generic-qemu/hd0.disk /my_trace ../../phoenix-rtos-kernel/perf/tsdl/metadata out.pftrace