On 03/02/2021 17:18, Ian Jackson wrote:
> Andrew Cooper writes ("[PATCH for-4.15 2/2] libs/foreignmem: Fix/simplify
> errno handling for map_resource"):
>> Simplify the FreeBSD logic, and duplicate it for NetBSD as the userpace ABI
>> appears to be to consistently provide EOPNOTSUPP for missing Xen/Kernel
>> support.
>>
>> The Linux logic was contorted in what appears to be a deliberate attempt to
>> skip the now-deleted logic for the EOPNOTSUPP case. Simplify it.
> AFAICT this is a mixture of cleanup/refactoring, and bugfix. Is that
> correct ?
>
>> diff --git a/tools/libs/foreignmemory/netbsd.c
>> b/tools/libs/foreignmemory/netbsd.c
>> index 565682e064..c0b1b8f79d 100644
>> --- a/tools/libs/foreignmemory/netbsd.c
>> +++ b/tools/libs/foreignmemory/netbsd.c
>> @@ -147,6 +147,9 @@ int osdep_xenforeignmemory_map_resource(
>> rc = ioctl(fmem->fd, IOCTL_PRIVCMD_MMAP_RESOURCE, &mr);
>> if ( rc )
>> {
>> + if ( errno == ENOSYS )
>> + errno = EOPNOTSUPP;
>> +
>> if ( fres->addr )
>> {
>> int saved_errno = errno;
> Specifically, I guess this is the bugfix ?
It is a bugfix on NetBSD (for brand new functionality), and cleanup on
FreeBSD/Linux specifically split out of the previous patch.
~Andrew
