Messages in this thread | | | Date | Wed, 28 Jan 2009 09:48:21 -0800 | From | "H. Peter Anvin" <> | Subject | Re: [: [git pull] headers_check fixes] |
| |
Arnd Bergmann wrote: > > For the specific x86 swab code, that would certainly be the simplest > way, user space should not be using those inline assemblies either > way. > > I think the more interesting question is whether we want to export > *any* inline helpers that are not part of the ABI to user space. > We already killed most of them (spinlocks, atomics, ...) and what > remains is basically just the byteorder code. All that is required > for the ABI is the information whether the system is big- or > little-endian, but not all the rest. >
In general, no. The byteswap API is a legacy exception.
-hpa
| |