Hi,
[taking Mark Kettenis into CC to hear his side of the PCI access
granularity debate]
On 29.06.2010 13:14, Martin Mares wrote:
>> Fix pciutils non-32bit PCI write on OpenBSD.
>> OpenBSD supports only 32bit accesses to PCI config space. Emulate
>> smaller writes with a read/modify/write cycle.
Hello!
> Fix pciutils non-32bit PCI write on OpenBSD.
> OpenBSD supports only 32bit accesses to PCI config space. Emulate
> smaller writes with a read/modify/write cycle.
Unfortunately, this is wrong ... PCI config space writes can have
arbitrary side-effects (even reads can, sigh), even if you w
On 2010/06/29 10:29, Carl-Daniel Hailfinger wrote:
> Fix pciutils non-32bit PCI write on OpenBSD.
> OpenBSD supports only 32bit accesses to PCI config space. Emulate
> smaller writes with a read/modify/write cycle.
>
> First try, untested. May break on bigendian architectures (actually, the
> exis
Fix pciutils non-32bit PCI write on OpenBSD.
OpenBSD supports only 32bit accesses to PCI config space. Emulate
smaller writes with a read/modify/write cycle.
First try, untested. May break on bigendian architectures (actually, the
existing code looks like it is broken on big endian for either read
