Re: [Qemu-devel] [PATCH 08/20] memory: store MemoryRegionSection pointers in phys_map

[Peter Maydell]

On 14 February 2012 09:27, Avi Kivity <address@hidden> wrote:
> Instead of storing PhysPageDesc, store pointers to MemoryRegionSections.
> The various offsets (phys_offset & ~TARGET_PAGE_MASK,
> PHYS_OFFSET & TARGET_PAGE_MASK, region_offset) can all be synthesized
> from the information in a MemoryRegionSection.  Adjust phys_page_find()
> to synthesize a PhysPageDesc.

git bisect blames this commit (5312bd8b3) for causing a Linux kernel
on spitz to produce a bunch of pxa2xx_i2c warnings that weren't
being emitted before:

$ ./arm-softmmu/qemu-system-arm -M spitz --kernel ~/linaro/zaurus/zImage
spitz_out_switch: Charging off.
pxa2xx_i2c_read: Bad register 0xffffff90
pxa2xx_i2c_write: Bad register 0xffffff90
pxa2xx_i2c_write: Bad register 0xffffff98
pxa2xx_i2c_read: Bad register 0xffffff90
pxa2xx_i2c_write: Bad register 0xffffff90
pxa2xx_i2c_write: Bad register 0xffffffa0
pxa2xx_i2c_write: Bad register 0xffffff90
pxa2xx_i2c_read: Bad register 0xffffff90
pxa2xx_i2c_write: Bad register 0xffffff90
spitz_lcdtg_transfer: LCD in VGA mode
spitz_bl_update: LCD Backlight now at 47/63
spitz_out_switch: Charging on.

The 'spitz' messages are just chatter, but the pxa_i2c 'Bad register'
warnings are new with this commit.

You can get the kernel image from here:
https://bugs.launchpad.net/qemu-linaro/+bug/885239/+attachment/2583099/+files/zImage

NB that the kernel will panic because it can't find init because
we haven't given it a rootfs, but that's not necessary for the
purposes of this regression.

I haven't looked any closer at what's going on here yet, I've just
run the git bisect on it...

-- PMM