Well, this is to allow a retry in case the cache wasn't properly flushed
already. This came about because the blrl instruction in the GOT was seen
as an illegal instruction, and it didn't seem to matter if I synced in
pmap_enter(), it would still SIGILL probabilistically. Since this block
was alr
Is this reasonable? What if the junk in the cache happened to be a
*valid* instruction? Won't this approach result in silent corruption and
later failure?
-Nathan
On 11/19/18 3:54 PM, Justin Hibbits wrote:
> Author: jhibbits
> Date: Mon Nov 19 23:54:49 2018
> New Revision: 340653
> URL: https://sv
Author: jhibbits
Date: Mon Nov 19 23:54:49 2018
New Revision: 340653
URL: https://svnweb.freebsd.org/changeset/base/340653
Log:
powerpc: Sync icache on SIGILL, in case of cache issues
The update of jemalloc to 5.1.0 exposed a cache syncing issue on a Freescale
e500 base system. There was
