126928 - PR_SetFDInheritable not supported on OS/2

Status: RESOLVED FIXED

(NSPR :: NSPR, defect, P3)

Description

[Julien Pierre]

This is one of many NSPR functions that fail on OS/2 today.

This function returns -5994 on OS/2. This was found when running selfserv from
NSS. This function is just a stub and not implemented.
I believe that there is no permission context in OS/2 for sockets and that all
sockets are automatically in the inheritable state, and that any fd can be open
by any process (I know, that's not secure) by socket number. I think that was
true of the 16-bit TCP/IP stacks that IBM had in Warp 4.0, but I'm not sure
about the new 32-bit TCP/IP stacks in Warp Server for e-business. Mike, it would
be nice to have  an opinion on the right implementation. Maybe we don't need to
do anything and just return success.

Comment 1

[Mike Kaply [:mkaply]]

Achim, do you or does anyone on your team have an idea on this one?

Comment 2

[Mike Kaply [:mkaply]]

switch (fd->methods->file_type) { case PR_DESC_SOCKET_TCP: case
PR_DESC_SOCKET_UDP: return PR_SUCCESS; default: <old code> }

Comment 3

[knut st. osmundsen]

This is entirely broken.
DosSetFHState is called with wrong flags (you have to mask off part of the query
result).

Comment 4

[knut st. osmundsen]

Attachment: Suggested fix.

Comment 5

[timeless]

Comment on attachment 128173 [details] [diff] [review]
Suggested fix.

> +            /* Mask off flags DosSetFHState don't want. */
don't => doesn't

Comment 6

[Mike Kaply [:mkaply]]

Comment on attachment 128173 [details] [diff] [review]
Suggested fix.

Please use APIRET rc instead of int rv and use NO_ERROR instead of 0.

Thanks!

Comment 7

[Mike Kaply [:mkaply]]

wtc, I've checked this in on the branch. Please check in on NSPR trunk or give
me authority :)

Comment 8

[Wan-Teh Chang]

Patch checked in on NSPR trunk (NSPR 4.5).