]> xenbits.xensource.com Git - qemu-upstream-4.2-testing.git/commit
cpu_physical_memory_write_rom() needs to do TB invalidates
authorDavid Gibson <david@gibson.dropbear.id.au>
Thu, 21 Feb 2013 12:16:41 +0000 (12:16 +0000)
committerStefano Stabellini <stefano.stabellini@eu.citrix.com>
Thu, 21 Feb 2013 12:16:41 +0000 (12:16 +0000)
commit321fa7c50b755080964008a03a894d42ff3bdba4
treeedac3085e21737632b9a4acbf491c736478c084e
parent4fb3a19e874efb97ae8897764a56c2da833bde80
cpu_physical_memory_write_rom() needs to do TB invalidates

cpu_physical_memory_write_rom(), despite the name, can also be used to
write images into RAM - and will often be used that way if the machine
uses load_image_targphys() into RAM addresses.

However, cpu_physical_memory_write_rom(), unlike cpu_physical_memory_rw()
doesn't invalidate any cached TBs which might be affected by the region
written.

This was breaking reset (under full emu) on the pseries machine - we loaded
our firmware image into RAM, and while executing it rewrite the code at
the entry point (correctly causing a TB invalidate/refresh).  When we
reset the firmware image was reloaded, but the TB from the rewrite was
still active and caused us to get an illegal instruction trap.

This patch fixes the bug by duplicating the tb invalidate code from
cpu_physical_memory_rw() in cpu_physical_memory_write_rom().

Backport of original commit 0b57e287138728f72d88b06e69b970c5d745c44a

Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Signed-off-by: Alex Bligh <alex@alex.org.uk>
exec.c