kernel/drivers/net/ethernet/intel/fm10k, branch linux-4.3.y

mm: make page pfmemalloc check more robust

2015-08-21T21:30:10Z

Commit c48a11c7ad26 ("netvm: propagate page->pfmemalloc to skb") added checks for page->pfmemalloc to __skb_fill_page_desc(): if (page->pfmemalloc && !page->mapping) skb->pfmemalloc = true; It assumes page->mapping == NULL implies that page->pfmemalloc can be trusted. However, __delete_from_page_cache() can set set page->mapping to NULL and leave page->index value alone. Due to being in union, a non-zero page->index will be interpreted as true page->pfmemalloc. So the assumption is invalid if the networking code can see such a page. And it seems it can. We have encountered this with a NFS over loopback setup when such a page is attached to a new skbuf. There is no copying going on in this case so the page confuses __skb_fill_page_desc which interprets the index as pfmemalloc flag and the network stack drops packets that have been allocated using the reserves unless they are to be queued on sockets handling the swapping which is the case here and that leads to hangs when the nfs client waits for a response from the server which has been dropped and thus never arrive. The struct page is already heavily packed so rather than finding another hole to put it in, let's do a trick instead. We can reuse the index again but define it to an impossible value (-1UL). This is the page index so it should never see the value that large. Replace all direct users of page->pfmemalloc by page_is_pfmemalloc which will hide this nastiness from unspoiled eyes. The information will get lost if somebody wants to use page->index obviously but that was the case before and the original code expected that the information should be persisted somewhere else if that is really needed (e.g. what SLAB and SLUB do). [akpm@linux-foundation.org: fix blooper in slub] Fixes: c48a11c7ad26 ("netvm: propagate page->pfmemalloc to skb") Signed-off-by: Michal Hocko Debugged-by: Vlastimil Babka Debugged-by: Jiri Bohac Cc: Eric Dumazet Cc: David Miller Acked-by: Mel Gorman Cc: [3.6+] Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds

fm10k: Fix missing braces after if statement

2015-06-17T21:21:41Z

While reviewing the code I noticed that one of the commits added an if statement followed by a for loop, but the if statement was missing the braces around the loop. This change corrects the coding style error. Signed-off-by: Alexander Duyck Acked-by: Jacob Keller Tested-by: Krishneil Singh Signed-off-by: Jeff Kirsher

fm10k: fix iov_msg_lport_state_pf issue

2015-06-17T21:21:36Z

When a VF issues an LPORT_STATE request to enable a port that is already enabled, the PF will first disable the VF LPORT. Then it should re-enable the VF again with the new requested settings. This ensures that any switch rules are cleared by deleting the LPORT on the switch. However, the flow is bugged because we actually check if the VF is enabled at the end, and thus don't re-enable it. Fix the flow so that we actually clear the enabled flags as part of our removal of the LPORT. Signed-off-by: Jacob Keller Tested-by: Krishneil Singh Signed-off-by: Jeff Kirsher

fm10k: remove err_no reference in fm10k_mbx.c

2015-06-17T21:21:30Z

The reference to err_no was left around after a previous code refactor. We never use the value, and it doesn't seem to be used in side a hidden macro reference. Discovered via cppcheck. Signed-off-by: Jacob Keller Tested-by: Krishneil Singh Signed-off-by: Jeff Kirsher

fm10k: fix incorrect DIR_NEVATIVE bit in 1588 code

2015-06-17T21:21:26Z

The SYSTIME_CFG.Adjust Direction bit is actually supposed to indicate that the adjustment is positive. Fix the code to align correctly with hardware and documentation. Signed-off-by: Jacob Keller Tested-by: Krishneil Singh Signed-off-by: Jeff Kirsher

fm10k: pack TLV overlay structures

2015-06-17T21:21:22Z

This patch adds the __attribute__((packed)) indicator to some structures which are overlayed onto a TLV message. These structures must be packed as small as possible in order to correctly align when copied into the mailbox buffer. Without doing so, the receiving mailbox code incorrectly parses the values and we get invalid message responses from the switch manager software. Signed-off-by: Jacob Keller Tested-by: Krishneil Singh Signed-off-by: Jeff Kirsher

fm10k: re-map all possible VF queues after a VFLR

2015-06-17T21:21:18Z

During initialization, the VF counts its rings by walking the TQDLOC registers. This works only if the TQMAP/RQMAP registers are set to map all of the out-of-bound rings back to the first one. This allows the VF to cleanly detect when it has run out of queues. Update the PF code so that it resets the empty TQMAP/RQMAP registers post-VFLR to prevent innocent VF drivers from triggering malicious driver events. Signed-off-by: Matthew Vick Signed-off-by: Jacob Keller Tested-by: Krishneil Singh Signed-off-by: Jeff Kirsher

fm10k: force LPORT delete when updating VLAN or MAC address

2015-06-17T21:21:15Z

Currently, we don't notify the switch at all when the PF administratively sets a new VLAN or MAC address. This causes the old addresses to remain valid on the switch table. Since the PF is overriding any configuration done directly by the VF, we choose to simply re-create the LPORT for the VF. This does mean that all rules for the VF will be dropped when we set something directly via the PF, but it prevents some weird issues where the MAC/VLAN table retains some stale configuration. Signed-off-by: Jacob Keller Tested-by: Krishneil Singh Signed-off-by: Jeff Kirsher

fm10k: use dma_set_mask_and_coherent in fm10k_probe

2015-06-17T21:21:10Z

This patch cleans up the use of dma_get_required_mask and uses the simpler dma_set_mask_and_coherent function instead of doing these as separate steps. I removed the dma_get_required_mask call because based on some minimal testing it appears that either (a) we're not doing the right thing with the call or (b) we don't need it anyways. If the value returned is <48bits, we'll end up trying with 48 bits anyways. If it's over 48bits, fm10k can't support that anyways, and we should try 48bits. If 48bits fails, we'll fallback to 32bits. This cleans up some very funky code. Signed-off-by: Jacob Keller Signed-off-by: Jeff Kirsher

fm10k: trivial fixup message style to include a colon

2015-06-17T21:21:01Z

Also use %d for error values, since printing in hexadecimal is probably not helpful. Signed-off-by: Jacob Keller Signed-off-by: Jeff Kirsher