- 06 Jun, 2017 10 commits
-
-
Benjamin Poirier authored
Some statistics passed to ethtool are garbage because e1000e_get_stats64() doesn't write them, for example: tx_heartbeat_errors. This leaks kernel memory to userspace and confuses users. Do like ixgbe and use dev_get_stats() which first zeroes out rtnl_link_stats64. Fixes: 5944701d ("net: remove useless memset's in drivers get_stats64") Reported-by:
Stefan Priebe <s.priebe@profihost.ag> Signed-off-by:
Benjamin Poirier <bpoirier@suse.com> Tested-by:
Aaron Brown <aaron.f.brown@intel.com> Signed-off-by:
Jeff Kirsher <jeffrey.t.kirsher@intel.com>
-
Benjamin Poirier authored
Given that all callers of igb_update_stats() pass the same two arguments: (adapter, &adapter->stats64), the second argument can be removed. Signed-off-by:
-
Jacob Keller authored
The igb driver has logic to handle only one Tx timestamp at a time, using a state bit lock to avoid multiple requests at once. It may be possible, if incredibly unlikely, that a Tx timestamp event is requested but never completes. Since we use an interrupt scheme to determine when the Tx timestamp occurred we would never clear the state bit in this case. Add an igb_ptp_tx_hang() function similar to the already existing igb_ptp_rx_hang() function. This function runs in the watchdog routine and makes sure we eventually recover from this case instead of permanently disabling Tx timestamps. Note: there is no currently known way to cause this without hacking the driver code to force it. Signed-off-by:
Jacob Keller <jacob.e.keller@intel.com> Tested-by:
-
Jacob Keller authored
The igb driver can only handle one Tx timestamp request at a time. This means it is possible for an application timestamp request to be ignored. There is no easy way for an administrator to determine if this occurred. Add a new statistic which tracks this, tx_hwtstamp_skipped. Signed-off-by:
-
Jacob Keller authored
The e1000e driver can only handle one Tx timestamp request at a time. This means it is possible for an application timestamp request to be ignored. There is no easy way for an administrator to determine if this occurred. Add a new statistic which tracks this, tx_hwtstamp_skipped. Signed-off-by:
-
Jacob Keller authored
The igb driver uses a state bit lock to avoid handling more than one Tx timestamp request at once. This is required because hardware is limited to a single set of registers for Tx timestamps. The state bit lock is not properly cleaned up during igb_xmit_frame_ring() if the transmit fails such as due to DMA or TSO failure. In some hardware this results in blocking timestamps until the service task times out. In other hardware this results in a permanent lock of the timestamp bit because we never receive an interrupt indicating the timestamp occurred, since indeed the packet was never transmitted. Fix this by checking for DMA and TSO errors in igb_xmit_frame_ring() and properly cleaning up after ourselves when these occur. Reported-by:
Reported-by: David Mirabito <davidm@metamako.com> Signed-off-by:
-
Jacob Keller authored
Hardware related to the igb driver has a limitation of only handling one Tx timestamp at a time. Thus, the driver uses a state bit lock to enforce that only one timestamp request is honored at a time. Unfortunately this suffers from a simple race condition. The bit lock is not cleared until after skb_tstamp_tx() is called notifying the stack of a new Tx timestamp. Even a well behaved application which sends only one timestamp request at once and waits for a response might wake up and send a new packet before the bit lock is cleared. This results in needlessly dropping some Tx timestamp requests. We can fix this by unlocking the state bit as soon as we read the Timestamp register, as this is the first point at which it is safe to unlock. To avoid issues with the skb pointer, we'll use a copy of the pointer and set the global variable in the driver structure to NULL first. This ensures that the next timestamp request does not modify our local copy of the skb pointer. This ensures that well behaved applications do not accidentally race with the unlock bit. Obviously an application which sends multiple Tx timestamp requests at once will still only timestamp one packet at a time. Unfortunately there is nothing we can do about this. Reported-by:
-
Jacob Keller authored
The e1000e driver and related hardware has a limitation on Tx PTP packets which requires we limit to timestamping a single packet at once. We do this by verifying that we never request a new Tx timestamp while we still have a tx_hwtstamp_skb pointer. Unfortunately the driver suffers from a race condition around this. The tx_hwtstamp_skb pointer is not set to NULL until after skb_tstamp_tx() is called. This function notifies the stack and applications of a new timestamp. Even a well behaved application that only sends a new request when the first one is finished might be woken up and possibly send a packet before we can free the timestamp in the driver again. The result is that we needlessly ignore some Tx timestamp requests in this corner case. Fix this by assigning the tx_hwtstamp_skb pointer prior to calling skb_tstamp_tx() and use a temporary pointer to hold the timestamped skb until that function finishes. This ensures that the application is not woken up until the driver is ready to begin timestamping a new packet. This ensures that well behaved applications do not accidentally race with condition to skip Tx timestamps. Obviously an application which sends multiple Tx timestamp requests at once will still only timestamp one packet at a time. Unfortunately there is nothing we can do about this. Reported-by:
-
Arnd Bergmann authored
The new wake function is only used by the suspend/resume handlers that are defined in inside of an #ifdef, which can cause this harmless warning: drivers/net/ethernet/intel/igb/igb_main.c:7988:13: warning: 'igb_deliver_wake_packet' defined but not used [-Wunused-function] Removing the #ifdef, instead using a __maybe_unused annotation simplifies the code and avoids the warning. Fixes: b90fa876 ("igb: Enable reading of wake up packet") Signed-off-by:
Arnd Bergmann <arnd@arndb.de> Tested-by:
-
Matwey V Kornilov authored
The functions igb_read_phy_reg_gs40g/igb_write_phy_reg_gs40g (which were removed in 2a3cdead) explicitly selected the required page at every phy_reg access. Currently, igb_get_phy_id_82575 relays on the fact that page 0 is already selected. The assumption is not fulfilled for my Lex 3I380CW motherboard with integrated dual i211 based gigabit ethernet. This leads to igb initialization failure and network interfaces are not working: igb: Intel(R) Gigabit Ethernet Network Driver - version 5.4.0-k igb: Copyright (c) 2007-2014 Intel Corporation. igb: probe of 0000:01:00.0 failed with error -2 igb: probe of 0000:02:00.0 failed with error -2 In order to fix it, we explicitly select page 0 before first access to phy registers. See also: https://bugzilla.suse.com/show_bug.cgi?id=1009911 See also: http://www.lex.com.tw/products/pdf/3I380A&3I380CW.pdf Fixes: 2a3cdead ("igb: Remove GS40G specific defines/functions") Cc: <stable@vger.kernel.org> # 4.5+ Signed-off-by:
Matwey V Kornilov <matwey@sai.msu.ru> Tested-by:
-
- 05 Jun, 2017 30 commits
-
-
Colin Ian King authored
The u32 variable v is being checked to see if an error return is less than zero and this check has no effect because it is unsigned. Fix this by making v and int (this also matches the type of cb->bus_number which is assigned to the value in v). Detected by CoverityScan, CID#1440454 ("Unsigned compared against zero") Signed-off-by:Colin Ian King <colin.king@canonical.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Icenowy Zheng authored
The EPHY may be already enabled by bootloaders which have Ethernet capability (e.g. current U-Boot). Thus it should be reseted properly before doing the enabling sequence in the dwmac-sun8i driver, otherwise the EMAC reset process may fail if no cable is plugged, and then fail the dwmac-sun8i probing. Tested on Orange Pi PC, One and Zero. All the boards fail to have dwmac-sun8i probed with "EMAC reset timeout" without cable plugged before, and with this fix they're now all able to successfully probe the EMAC without cable plugged and then use the connection after a cable is hot-plugged in. Fixes: 9f93ac8d ("net-next: stmmac: Add dwmac-sun8i") Signed-off-by:
Icenowy Zheng <icenowy@aosc.io> Tested-by:
Corentin Labbe <clabbe.montjoie@gmail.com> Acked-by:
-
yuval.shaia@oracle.com authored
Make return value void since function never returns meaningfull value. Signed-off-by:
Yuval Shaia <yuval.shaia@oracle.com> Signed-off-by:
-
yuval.shaia@oracle.com authored
Make return value void since functions never returns meaningfull value. Signed-off-by:
-
yuval.shaia@oracle.com authored
Make return value void since function never return meaningfull value Signed-off-by:
-
Jiri Pirko authored
It really makes no sense to have cls_act enabled without cls. In that case, the cls_act code is dead. So select it. This also fixes an issue recently reported by kbuild robot: [linux-next:master 1326/4151] net/sched/act_api.c:37:18: error: implicit declaration of function 'tcf_chain_get' Reported-by:
kbuild test robot <fengguang.wu@intel.com> Fixes: db50514f ("net: sched: add termination action to allow goto chain") Signed-off-by:
Jiri Pirko <jiri@mellanox.com> Signed-off-by:
-
Rosen, Rami authored
commit d91824c0 ("genetlink: register family ops as array") removed the ops_list member from both genl_family and genl_ops; while the documentation of genl_family was updated accordingly by this patch, ops_list remained in the documentation of the genl_ops object. This patch fixes it by removing ops_list from genl_ops documentation. Signed-off-by:
Rami Rosen <rami.rosen@intel.com> Signed-off-by:
-
David S. Miller authored
Jiri Pirko says: ==================== mlxsw: Minor cleanup Fix small issues I noticed during the refactoring. First patch adds file name comments in the header file to make it clear what goes where. Second patch fixes a typo and third patch simply aligns RIF index allocation with similar allocations in the driver. ==================== Signed-off-by: -
Ido Schimmel authored
The way we usually allocate an index is by letting the allocation function return an error instead of an invalid index. Do the same for RIF index. Signed-off-by:
Ido Schimmel <idosch@mellanox.com> Reviewed-by:
Petr Machata <petrm@mellanox.com> Signed-off-by:
-
Ido Schimmel authored
Signed-off-by:
-
Ido Schimmel authored
Make it clear where functions are defined and move misplaced declaration to their correct place. Signed-off-by:
-
Yotam Gigi authored
Change the firmware file name to be in "mellanox" directory. This commit is a followup to the linux-firmware commit a4c72696f5f4 ("Mellanox: Add firmware for mlxsw_spectrum") Signed-off-by:Yotam Gigi <yotamg@mellanox.com> Reviewed-by:
-
David S. Miller authored
Yuval Mintz says: qed*: Support VF XDP attachment ==================== Each driver queue [Rx, Tx, XDP-forwarding] requires an allocated HW/FW connection + configured queue-zone. VF handling by the PF has several limitations that prevented adding the capability to perform XDP at driver-level: - The VF assumes there's 1-to-1 correspondance between the VF queue and the used connection, meaning q<x> is always going to use cid<x>, whereas for its own queues the PF is acquiring a new cid per each new queue. - There's a 1-to-1 correspondate between the VF-queues and the HW queue zones. While this is necessary for Rx-queues [as the queue-zone contains the producer], transmission queues can share the underlaying queue-zone [only shared configuration is coalescing]. But all VF<->PF communication mechanisms assume there's a single identifier that identify a queue [as queue-zone == queue], while sharing queue-zones requires passing additional information. - VFs currently don't try mapping a doorbell bar - there's a small doorbell window in the regview allowing VFs to doorbell up to 16 connections; but this window isn's wide enough for the added XDP forwarding queues. This series is going to add the necessary infrastrucutre to finally let our VFs support XDP assuming both the PF and VF drivers are sufficiently new [Legacy support would be retained both for older VFs and older PFs, but both will be needed for this new support to work]. Basically, the various database driver maintains for its queue-cids would be revised, and queue-cids would be identified using the (queue-zone, unique index) pair. The TLV mechanism would then be extended to allow VFs to communicate that unique-index as well as the already provided queue-zone. Finally, the VFs would try to map their doorbell bar and inform their PF that they're using it. Almost all the changes are in qed, with exception of #3 [which does some cleanup in qede as well] and #11 that actually enables the feature. ==================== Signed-off-by: -
Mintz, Yuval authored
This introduces 2 changes needed for XDP to be supported for VFs: a. On VF-side, publish the NDO based on qed outputs b. On PF-side, request qed to allocate sufficient cids per-VF to allow the child vfs to support it Signed-off-by:Yuval Mintz <Yuval.Mintz@cavium.com> Signed-off-by:
-
Mintz, Yuval authored
The final addition on the qed front - - VFs would now require their PFs to provide multiple CIDs - Based on the availability of connections from PF, determine whether XDP is feasible and share it with qede via dev_info. Signed-off-by:
-
Mintz, Yuval authored
VFs are currently not mapping their doorbell bar, instead relying on the small doorbell window they have in their limited regview bar. In order to increase the number of possible Tx connections [queues] employeed by VF past 16, we need to start using the doorbell bar if one such is exposed - VF would communicate this fact to PF which would return the size-bar internally configured into chip, according to which the VF would decide whether to actually utilize the doorbell bar. Signed-off-by:
-
Mintz, Yuval authored
This adds the infrastructure for supporting VFs that want to open multiple transmission queues on the same queue-zone. At this point, there are no VFs that actually request this functionality, but later patches would remedy that. a. VF and PF would communicate the capability during ACQUIRE; Legacy VFs would continue on behaving as they do today b. PF would communicate number of supported CIDs to the VF and would enforce said limitation c. Whenever VF passes a request for a given queue configuration it would also pass an associated index within said queue-zone Signed-off-by: -
Mintz, Yuval authored
Allow the infrastructure a PF maintains for each one of its VFs to support multiple queue-cids on a single queue-zone. Signed-off-by:
-
Mintz, Yuval authored
Until now we used to have a single VF legacy compatibility mode, one that affected the place of the Rx producers of those VFs [mostly]. As PF would soon support allocating CIDs for VFs instead of having a static CID<->queue configuration for them, we'll need to have an additional legacy mode since existing VFs would need to continue on using the older mode of operation. Change the infrastrucutre so that the legacy would be able to indicate which of the legacy behaviors is needed for a given VF. Signed-off-by:
-
Mintz, Yuval authored
When a queue-cid is allocated, assign an index inside that's CID's queue-zone. For PFs and VFS, this number is going to be unique and derive from a per-queue-zone bitmap, while for PF's VFs queues the number is currently going to constant; Later, we'd add the capability of a VF to communicate such an index to its PF. Signed-off-by:
-
Mintz, Yuval authored
We're going to need additional information for queue-cids that a PF creates for its VFs, so start by refactoring existing logic used for initializing said struct into receiving a structure encapsulating the VF-specific information that needs to be provided. This also introduces QED_QUEUE_CID_SELF - each queue-cid would hold an indication to whether it belongs to the hw-function holding it [whether that's a PF or a VF], or else what's the VF id it belongs to. Signed-off-by:
-
Mintz, Yuval authored
Part of an effort of a cleaner seperation between qed and the protocol drivers, the L2 interface is to use the SB structure for initialization purposes opaquely. Signed-off-by:
-
Mintz, Yuval authored
First step in allowing a single PF/VF to open multiple queues on the same queue zone is to add per-hwfn database of queue-cids as a two-dimensional array where entry would be according to [queue zone][internal index]. Signed-off-by:
-
Mintz, Yuval authored
Each PF has a bitmap for its own ranges of CIDs, to allow easy grabbing of an available CID when such is needed. But VFs are not using the same mechanism, instead relying on hard-coded CIDs [ queue-index == cid ]. As an infrastructure step toward increasing number of CIDs of VFs, the PF is going to maintain bitmaps for the VF CIDs as well - the bitmaps would be per-VF and the ranges would be the same [in HW all VFs of a given PF have the same mapping of CIDs, and the HW is capable of distinguishing between those according to the VF index] Signed-off-by:
-
David S. Miller authored
Jason A. Donenfeld says: ==================== net: Avoiding stack overflow in skb_to_sgvec The recent bug with macsec and historical one with virtio have indicated that letting skb_to_sgvec trounce all over an sglist without checking the length is probably a bad idea. And it's not necessary either: an sglist already explicitly marks its last item, and the initialization functions are diligent in doing so. Thus there's a clear way of avoiding future overflows. So, this patchset, from a high level, makes skb_to_sgvec return a potential error code, and then adjusts all callers to check for the error code. There are two situations in which skb_to_sgvec might return such an error: 1) When the passed in sglist is too small; and 2) When the passed in skbuff is too deeply nested. So, the first patch in this series handles the issues with skb_to_sgvec directly, and the remaining ones then handle the call sites. ==================== Signed-off-by: -
Jason A. Donenfeld authored
Signed-off-by:
Jason A. Donenfeld <Jason@zx2c4.com> Reviewed-by:
Sergei Shtylyov <sergei.shtylyov@cogentembedded.com> Cc: "Michael S. Tsirkin" <mst@redhat.com> Cc: Jason Wang <jasowang@redhat.com> Signed-off-by:
-
Jason A. Donenfeld authored
Signed-off-by:
-
Jason A. Donenfeld authored
Signed-off-by:
David Howells <dhowells@redhat.com> Signed-off-by:
-
Jason A. Donenfeld authored
Signed-off-by:
-
Jason A. Donenfeld authored
This is a defense-in-depth measure in response to bugs like 4d6fa57b ("macsec: avoid heap overflow in skb_to_sgvec"). There's not only a potential overflow of sglist items, but also a stack overflow potential, so we fix this by limiting the amount of recursion this function is allowed to do. Not actually providing a bounded base case is a future disaster that we can easily avoid here. As a small matter of house keeping, we take this opportunity to move the documentation comment over the actual function the documentation is for. While this could be implemented by using an explicit stack of skbuffs, when implementing this, the function complexity increased considerably, and I don't think such complexity and bloat is actually worth it. So, instead I built this and tested it on x86, x86_64, ARM, ARM64, and MIPS, and measured the stack usage there. I also reverted the recent MIPS changes that give it a separate IRQ stack, so that I could experience some worst-case situations. I found that limiting it to 24 layers deep yielded a good stack usage with room for safety, as well as being much deeper than any driver actually ever creates. Signed-off-by:
-
