stack trace from sctp_getsockopt_local_addrs()

In sctp_getsockopt_local_addrs() we can potentially try allocate too much memory depending on what the user passes to us as "len".

net/sctp/socket.c
4489 to = optval + offsetof(struct sctp_getaddrs,addrs);
4490 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4491
4492 addrs = kmalloc(space_left, GFP_KERNEL);
4493 if (!addrs)
4494 return -ENOMEM;
4495

It's harmless. "space_left" is never zero and we handle memory allocation failures correctly but it's still a little messy and can lead to a stack trace. It's a WARN_ONCE() so it's not a huge deal.

Sep  6 20:57:07 bicker kernel: [84533.160286] ------------[ cut here ]------------
Sep  6 20:57:07 bicker kernel: [84533.160304] WARNING: at mm/page_alloc.c:1971 __alloc_pages_nodemask+0x153/0x4ed()
Sep  6 20:57:07 bicker kernel: [84533.160311] Hardware name: 900HA
Sep  6 20:57:07 bicker kernel: [84533.160315] Modules linked in: ufs ppp_deflate zlib_deflate bsd_comp ppp_async crc_ccitt pp
p_generic slhc sctp crc32c libcrc32c option usb_wwan usbserial ath5k atl1e ath
Sep  6 20:57:07 bicker kernel: [84533.160359] Pid: 8949, comm: a.out Not tainted 2.6.36-rc3+ #39
Sep  6 20:57:07 bicker kernel: [84533.160365] Call Trace:
Sep  6 20:57:07 bicker kernel: [84533.160380]  [<c103019e>] warn_slowpath_common+0x60/0x75
Sep  6 20:57:07 bicker kernel: [84533.160391]  [<c10301c2>] warn_slowpath_null+0xf/0x13
Sep  6 20:57:07 bicker kernel: [84533.160402]  [<c10914ff>] __alloc_pages_nodemask+0x153/0x4ed
Sep  6 20:57:07 bicker kernel: [84533.160414]  [<c10307de>] ? release_console_sem+0x177/0x1a4
Sep  6 20:57:07 bicker kernel: [84533.160428]  [<c10918a8>] __get_free_pages+0xf/0x21
Sep  6 20:57:07 bicker kernel: [84533.160440]  [<c10b1625>] __kmalloc+0x2f/0xf9
Sep  6 20:57:07 bicker kernel: [84533.160473]  [<f83964bc>] sctp_getsockopt+0x99d/0x16b5 [sctp]

This could happen if you called
getsockopt(sd, SOL_SCTP, SCTP_GET_LOCAL_ADDRS, getaddrs, &len);
without initializing "len".

regards,
dan carpenter