Re: [OpenBIOS] [Qemu-devel] [PATCH 01/27] Clean up PowerPC SLB handling code
QEMU HEAD still uses a 32-bit binary for both 32-bit and 64-bit. That one uses mtsrin so will need the compatibility, it seemed affected, too. OpenBIOS SVN HEAD (blob) uses slb* as linked to. We're in the preparation of 1.1 and I need to test it before we can update the QEMU binary. ;) Sorry for top-posting, Android sucks. Andreas David Gibson <david@gibson.dropbear.id.au> schrieb:
On Thu, May 19, 2011 at 08:00:53AM +0200, Andreas Färber wrote:
Am 19.05.2011 um 07:39 schrieb David Gibson:
On Thu, May 19, 2011 at 07:35:30AM +0200, Andreas Färber wrote:
Am 25.03.2011 um 04:21 schrieb David Gibson:
Currently the SLB information when emulating a PowerPC 970 is storeed in a structure with the unhelpfully named fields 'tmp' and 'tmp64'. While the layout in these fields does match the description of the SLB in the architecture document, it is not convenient either for looking up the SLB, or for emulating the slbmte instruction.
This patch, therefore, reorganizes the SLB entry structure to be divided in the the "ESID related" and "VSID related" fields as they are divided in instructions accessing the SLB.
In addition to making the code smaller and more readable, this will make it easier to implement for the 1TB segments used in more recent PowerPC chips.
Signed-off-by: David Gibson <dwg@au1.ibm.com>
According to my bisect, this patch broke ppc64 OpenBIOS.
David, would you please take a look?
Uh, sure. can you describe the symptoms of the breakage, and the exact qemu setup you've observed the problem with?
$ uname -a Darwin PMG5-3.local 9.8.0 Darwin Kernel Version 9.8.0: Wed Jul 15 16:57:01 PDT 2009; root:xnu-1228.15.4~1/RELEASE_PPC Power Macintosh # Mac OS X 10.5.8 on dual-core G5
$ gcc-4.2 --version powerpc-apple-darwin9-gcc-4.2.1 (GCC) 4.2.1 (Apple Inc. build 5577) Copyright (C) 2007 Free Software Foundation, Inc. This is free software; see the source for copying conditions. There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
../qemu/configure --prefix=/Users/andreas/QEMU/latest64 \ --target-list=ppc-softmmu,ppc64-softmmu \ --extra-cflags="-arch ppc64" --extra-ldflags="-arch ppc64" --cc=gcc-4.2 --host-cc=gcc-4.2 \ --enable-cocoa --disable-kvm --disable-strip --disable-docs --enable-io-thread $* # I/O thread doesn't matter
/Users/andreas/QEMU/qemu64/ppc64-softmmu/qemu-system-ppc64 -boot d -cdrom /Users/andreas/QEMU/AIX/dvd.1022A4_OBETA_710.iso \ -bios /Users/andreas/QEMU/OpenBIOS/openbios/obj-ppc64/openbios-qemu.elf -m 1024 \ -M mac99 -prom-env 'auto-boot?=false' -nographic
CD should be irrelevant, -boot c -hda /dev/null should work the same.
Um, yeah, I thought putting an AIX dvd into a Mac was a bit odd.
Expected: OpenBIOS banner and > prompt Observed: DSI exception on mtmsrd, no serial output
Does this happen only with your openbios image, or does it also happen with the one included in qemu?
My working theory is that openbios is using the old 32-64 bridge stuff where segment register accesses are emulated to populate certain SLB slots. I have a vague recollection that I saw the code implementing that when I was doing the cleanup and convinced myself it was not necessary. More information as I investigate further.
(The regression on ppc32 AIX observed by Kenneth seems to be a different issue, introduced earlier.)
I don't know what that regression is; I don't read qemu-devel exhaustively. If you want my attention, best to CC me directly.
-- David Gibson | I'll have my music baroque, and my code david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_ | _way_ _around_! http://www.ozlabs.org/~dgibson
On Thu, May 19, 2011 at 10:25:04AM +0200, Andreas Färber wrote:
QEMU HEAD still uses a 32-bit binary for both 32-bit and 64-bit. That one uses mtsrin so will need the compatibility, it seemed affected, too.
OpenBIOS SVN HEAD (blob) uses slb* as linked to. We're in the preparation of 1.1 and I need to test it before we can update the QEMU binary. ;)
Sorry for top-posting, Android sucks.
So, my theory was half right. It was a problem with 64-bit mtsr emulation, but it wasn't that I just removed that code with the SLB cleanup. The code was still there and *almost* right. I was off by one in one shift, causing the storage key bits to end up in the wrong place in the SLB entry. I'll send out the patch right after I've sent this mail. -- David Gibson | I'll have my music baroque, and my code david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_ | _way_ _around_! http://www.ozlabs.org/~dgibson
On 20.05.2011, at 05:34, David Gibson wrote:
On Thu, May 19, 2011 at 10:25:04AM +0200, Andreas Färber wrote:
QEMU HEAD still uses a 32-bit binary for both 32-bit and 64-bit. That one uses mtsrin so will need the compatibility, it seemed affected, too.
OpenBIOS SVN HEAD (blob) uses slb* as linked to. We're in the preparation of 1.1 and I need to test it before we can update the QEMU binary. ;)
Sorry for top-posting, Android sucks.
So, my theory was half right. It was a problem with 64-bit mtsr emulation, but it wasn't that I just removed that code with the SLB cleanup. The code was still there and *almost* right. I was off by one in one shift, causing the storage key bits to end up in the wrong place in the SLB entry. I'll send out the patch right after I've sent this mail.
Thanks a lot for tracking it down you two :) Alex
participants (3)
-
Alexander Graf -
Andreas Färber -
David Gibson