Don't access userspace directly from the kernel in nxge(4).

[brooks]

Summary

The previous code was obviously wrong, the new code makes the same
accesses to userspace via the correct mechanism. xge_ioctl_registers()
has further obvious issues, but there seems to be little point in fixing
them.

Test Plan

it compiles...

[jhb]

sys/dev/nxge/if_nxge.c

1377 ↗

(On Diff #40733)

If this is the only use, I would actually prefer to kill 'data' entirely to avoid someone trying to use additional bytes and instead do something like:

char query;
...

query = fubyte(ifreqp->ifr_data);
if (query == -1)
     return (EFAULT);

switch (query) {

Also, I wonder if we should prefer fuebyte() in new code instead? (Maybe kib@ has an opinion on that)

1547 ↗

(On Diff #40733)

The fact that it does copyout() already makes this function even more amusing.

[brooks]

sys/dev/nxge/if_nxge.c

1377 ↗

(On Diff #40733)

data is used later or I would have done that. We don't currently have fuebyte()

1547 ↗

(On Diff #40733)

The strcmp() cases will certainly not trigger in most cases because option is a char[2] followed by an int64_t so there's garbage in the pad unless the did a bzero first.

This driver is best fixed with svn rm, but I'm changing ifr_data use so needed to do a little fixing.

[jhb]

Agreed that the driver is dubious at best. I'm not sure if it's even really used. I don't think it was a commodity part.

sys/dev/nxge/if_nxge.c

1506 ↗

(On Diff #40733)

So this is the only place that uses 'data' and it is wrong (sizeof is wrong). I would probably fix this by not abusing data but just using a local 'char mode' and passing '&mode' to copyout().

[brooks]

• Split data into cmd and mode to be more clear what is going on.