| Message ID | 20101013191250.GA59919@triton8.kn-bremen.de |
|---|---|
| State | New |
| Headers | show |
On Wed, Oct 13, 2010 at 7:12 PM, Juergen Lock <qemu-l@jelal.kn-bremen.de> wrote: > The 2nd scoop's base address (0x08800040) now gets rounded down to > start of page which causes its io read/write callbacks to be passed > addresses 0x40 higher than the code expects: (as witnessed by > "Bad register offset" messages and failure to attach the internal > CF disk aka microdrive at least.) > > [There may be more bugs of this kind hiding in other targets, this > was just the one I tested...] The devices are passed an offset to base address. Perhaps the real problem is that scoop_init registers too much MMIO: 0x1000, when the real range should be only 0x28. Also the registers are in 4 byte intervals and any access to address between the registers also triggers a warning. What were the messages exactly?
--- a/hw/zaurus.c +++ b/hw/zaurus.c @@ -70,6 +70,10 @@ static uint32_t scoop_readb(void *opaque { ScoopInfo *s = (ScoopInfo *) opaque; + // XXX Workaround for base address (0x08800040 in this case) + // rounded down to start of page + addr &= 0x3f; + switch (addr) { case SCOOP_MCR: return s->mcr; @@ -104,6 +108,10 @@ static void scoop_writeb(void *opaque, t ScoopInfo *s = (ScoopInfo *) opaque; value &= 0xffff; + // XXX Workaround for base address (0x08800040 in this case) + // rounded down to start of page + addr &= 0x3f; + switch (addr) { case SCOOP_MCR: s->mcr = value;
The 2nd scoop's base address (0x08800040) now gets rounded down to start of page which causes its io read/write callbacks to be passed addresses 0x40 higher than the code expects: (as witnessed by "Bad register offset" messages and failure to attach the internal CF disk aka microdrive at least.) [There may be more bugs of this kind hiding in other targets, this was just the one I tested...] Signed-off-by: Juergen Lock <nox@jelal.kn-bremen.de>