Memory barriers and spin_unlock safety
A message that is not in the archive (sent off-list or not yet received): [email protected]
A message that is not in the archive (sent off-list or not yet received): [email protected]
A message that is not in the archive (sent off-list or not yet received): [email protected]
A message that is not in the archive (sent off-list or not yet received): [email protected]
A message that is not in the archive (sent off-list or not yet received): [email protected]
A message that is not in the archive (sent off-list or not yet received): [email protected]
Quoted 1 line · Linus Torvalds
Close, yes. HOWEVER, it's only really ordered wrt the "innermost" bus. I don't think PCI bridges are supposed to post PIO writes, but a x86 CPU basically won't stall for them forever. I _think_ they'll wait for it to hit that external bus, though.
PCI I/O writes are allowed to be posted by the host bus bridge (not other bridges) according to the PCI 2.2 spec. Maybe no x86 platform actually does this, but it's allowed, so technically a device would need to do a read in order to ensure that I/O writes have arrived at the device as well.
Signature · 3 lines
Robert Hancock Saskatoon, SK, Canada To email, remove "nospam" from [email protected] Home Page: http://www.roberthancock.com/
Quoted 5 lines · Robert Hancock
PCI I/O writes are allowed to be posted by the host bus bridge (not other bridges) according to the PCI 2.2 spec. Maybe no x86 platform actually does this, but it's allowed, so technically a device would need to do a read in order to ensure that I/O writes have arrived at the device as well.
Existing Linux drivers largely believe that PCI I/O cycles as opposed to MMIO cycles are not posted. At least one MIPS platform that did post them ended up ensuring PCI I/O cycle posting didn't occur to get a running Linux system - so its quite a deep assumption.