- Jan 14, 2021
-
-
Cristian Dumitrescu authored
Currently, the function i40e_construct_skb_zc only frees the input xdp buffer when the output skb is successfully built. On error, the function i40e_clean_rx_irq_zc does not commit anything for the current packet descriptor and simply exits the packet descriptor processing loop, with the plan to restart the processing of this descriptor on the next invocation. Therefore, on error the ring next-to-clean pointer should not advance, the xdp i.e. *bi buffer should not be freed and the current buffer info should not be invalidated by setting *bi to NULL. Therefore, the *bi should only be set to NULL when the function i40e_construct_skb_zc is successful, otherwise a NULL *bi will be dereferenced when the work for the current descriptor is eventually restarted. Fixes: 3b4f0b66 ("i40e, xsk: Migrate to new MEM_TYPE_XSK_BUFF_POOL") Signed-off-by:
Cristian Dumitrescu <cristian.dumitrescu@intel.com> Acked-by:
Björn Töpel <bjorn.topel@intel.com> Link: https://lore.kernel.org/r/20210111181138.49757-1-cristian.dumitrescu@intel.com Signed-off-by:
Jakub Kicinski <kuba@kernel.org>
-
- Dec 17, 2020
-
-
Sylwester Dziedziuch authored
When removing VFs for PF added to bridge there was an error I40E_AQ_RC_EINVAL. It was caused by not properly resetting and reinitializing PF when adding/removing VFs. Changed how reset is performed when adding/removing VFs to properly reinitialize PFs VSI. Fixes: fc60861e ("i40e: start up in VEPA mode by default") Signed-off-by:
Sylwester Dziedziuch <sylwesterx.dziedziuch@intel.com> Tested-by:
Konrad Jankowski <konrad0.jankowski@intel.com> Signed-off-by:
Tony Nguyen <anthony.l.nguyen@intel.com>
-
- Dec 16, 2020
-
-
Björn Töpel authored
On the Rx side, the next_to_use index points to the next item in the HW ring to be refilled/allocated, and next_to_clean points to the next item to potentially be processed. When the HW Rx ring is fully refilled, i.e. no packets has been processed, the next_to_use will be next_to_clean - 1. When the ring is fully processed next_to_clean will be equal to next_to_use. The latter case is where a bug is triggered. If the next_to_use bits are not cleared, and the "fully processed" state is entered, a stale descriptor can be processed. The skb-path correctly clear the status bit for the next_to_use descriptor, but the AF_XDP zero-copy path did not do that. This change adds the status bits clearing of the next_to_use descriptor. Fixes: 3b4f0b66 ("i40e, xsk: Migrate to new MEM_TYPE_XSK_BUFF_POOL") Signed-off-by:
-
- Dec 09, 2020
-
-
Björn Töpel authored
The page recycle code, incorrectly, relied on that a page fragment could not be freed inside xdp_do_redirect(). This assumption leads to that page fragments that are used by the stack/XDP redirect can be reused and overwritten. To avoid this, store the page count prior invoking xdp_do_redirect(). Longer explanation: Intel NICs have a recycle mechanism. The main idea is that a page is split into two parts. One part is owned by the driver, one part might be owned by someone else, such as the stack. t0: Page is allocated, and put on the Rx ring +--------------- used by NIC ->| upper buffer (rx_buffer) +--------------- | lower buffer +--------------- page count == USHRT_MAX rx_buffer->pagecnt_bias == USHRT_MAX t1: Buffer is received, and passed to the stack (e.g.) +--------------- | upper buff (skb) +--------------- used by NIC ->| lower buffer (rx_buffer) +--------------- page count == USHRT_MAX rx_buffer->pagecnt_bias == USHRT_MAX - 1 t2: Buffer is received, and redirected +--------------- | upper buff (skb) +--------------- used by NIC ->| lower buffer (rx_buffer) +--------------- Now, prior calling xdp_do_redirect(): page count == USHRT_MAX rx_buffer->pagecnt_bias == USHRT_MAX - 2 This means that buffer *cannot* be flipped/reused, because the skb is still using it. The problem arises when xdp_do_redirect() actually frees the segment. Then we get: page count == USHRT_MAX - 1 rx_buffer->pagecnt_bias == USHRT_MAX - 2 From a recycle perspective, the buffer can be flipped and reused, which means that the skb data area is passed to the Rx HW ring! To work around this, the page count is stored prior calling xdp_do_redirect(). Note that this is not optimal, since the NIC could actually reuse the "lower buffer" again. However, then we need to track whether XDP_REDIRECT consumed the buffer or not. Fixes: d9314c47 ("i40e: add support for XDP_REDIRECT") Reported-and-analyzed-by:Li RongQing <lirongqing@baidu.com> Signed-off-by:
George Kuruvinakunnel <george.kuruvinakunnel@intel.com> Signed-off-by:
-
- Nov 30, 2020
-
-
Björn Töpel authored
Add napi_id to the xdp_rxq_info structure, and make sure the XDP socket pick up the napi_id in the Rx path. The napi_id is used to find the corresponding NAPI structure for socket busy polling. Signed-off-by:
Daniel Borkmann <daniel@iogearbox.net> Acked-by:
Ilias Apalodimas <ilias.apalodimas@linaro.org> Acked-by:
Michael S. Tsirkin <mst@redhat.com> Acked-by:
Tariq Toukan <tariqt@nvidia.com> Link: https://lore.kernel.org/bpf/20201130185205.196029-7-bjorn.topel@gmail.com
-
- Nov 24, 2020
-
-
Stefan Assmann authored
When the virtual link state was set to "enable" ethtool would report link speed as 40000Mb/s regardless of the underlying device. Report the correct link speed. Example from a XXV710 NIC. Before: $ ip link set ens3f0 vf 0 state auto $ ethtool enp8s2 | grep Speed Speed: 25000Mb/s $ ip link set ens3f0 vf 0 state enable $ ethtool enp8s2 | grep Speed Speed: 40000Mb/s After: $ ip link set ens3f0 vf 0 state auto $ ethtool enp8s2 | grep Speed Speed: 25000Mb/s $ ip link set ens3f0 vf 0 state enable $ ethtool enp8s2 | grep Speed Speed: 25000Mb/s Signed-off-by:Stefan Assmann <sassmann@kpanic.de> Tested-by:
Aaron Brown <aaron.f.brown@intel.com> Signed-off-by:
-
Marek Majtyka authored
Remove a redundant assignment of the software ring pointer in the i40e driver. The variable is assigned twice with no use in between, so just get rid of the first occurrence. Fixes: 3b4f0b66 ("i40e, xsk: Migrate to new MEM_TYPE_XSK_BUFF_POOL") Signed-off-by:
Marek Majtyka <marekx.majtyka@intel.com> Acked-by:
-
Sylwester Dziedziuch authored
Prevent VFs from resetting when PF driver is being unloaded: - introduce new pf state: __I40E_VF_RESETS_DISABLED; - check if pf state has __I40E_VF_RESETS_DISABLED state set, if so, disable any further VFLR event notifications; - when i40e_remove (rmmod i40e) is called, disable any resets on the VFs; Previously if there were bare-metal VFs passing traffic and PF driver was removed, there was a possibility of VFs triggering a Tx timeout right before iavf_remove. This was causing iavf_close to not be called because there is a check in the beginning of iavf_remove that bails out early if adapter->state < IAVF_DOWN_PENDING. This makes it so some resources do not get cleaned up. Fixes: 6a9ddb36 ("i40e: disable IOV before freeing resources") Signed-off-by:
Slawomir Laba <slawomirx.laba@intel.com> Signed-off-by:
Brett Creeley <brett.creeley@intel.com> Signed-off-by:
-
- Nov 17, 2020
-
-
Magnus Karlsson authored
Use the new batched xsk interfaces for the Tx path in the i40e driver to improve performance. On my machine, this yields a throughput increase of 4% for the l2fwd sample app in xdpsock. If we instead just look at the Tx part, this patch set increases throughput with above 20% for Tx. Note that I had to explicitly loop unroll the inner loop to get to this performance level, by using a pragma. It is honored by both clang and gcc and should be ignored by versions that do not support it. Using the -funroll-loops compiler command line switch on the source file resulted in a loop unrolling on a higher level that lead to a performance decrease instead of an increase. Signed-off-by:
Magnus Karlsson <magnus.karlsson@intel.com> Signed-off-by:
John Fastabend <john.fastabend@gmail.com> Link: https://lore.kernel.org/bpf/1605525167-14450-6-git-send-email-magnus.karlsson@gmail.com
-
Magnus Karlsson authored
Remove the unnecessary access to the software ring for the AF_XDP zero-copy driver. This was used to record the length of the packet so that the driver Tx completion code could sum this up to produce the total bytes sent. This is now performed during the transmission of the packet, so no need to record this in the software ring. Signed-off-by:
-
- Nov 10, 2020
-
-
Dan Carpenter authored
The "failure" variable is used without being initialized. It should be set to false. Fixes: 8cbf7414 ("i40e, xsk: move buffer allocation out of the Rx processing loop") Signed-off-by:
Dan Carpenter <dan.carpenter@oracle.com> Acked-by:
-
Slawomir Laba authored
Fix MAC setting flow for the PF driver. Update the unicast VF's MAC address in VF structure if it is a new setting in i40e_vc_add_mac_addr_msg. When unicast MAC address gets deleted, record that and set the new unicast MAC address that is already waiting in the filter list. This logic is based on the order of messages arriving to the PF driver. Without this change the MAC address setting was interpreted incorrectly in the following use cases: 1) Print incorrect VF MAC or zero MAC ip link show dev $pf 2) Don't preserve MAC between driver reload rmmod iavf; modprobe iavf 3) Update VF MAC when macvlan was set ip link add link $vf address $mac $vf.1 type macvlan 4) Failed to update mac address when VF was trusted ip link set dev $vf address $mac This includes all other configurations including above commands. Fixes: f657a6e1 ("i40e: Fix VF driver MAC address configuration") Signed-off-by:
-
- Oct 14, 2020
-
-
Jaroslaw Gawin authored
Starting with API version 1.10 firmware for X722 devices has ability to change FEC settings in PHY. Code added in this patch allows changing FEC settings if the capability flag indicates the device supports this feature. Signed-off-by:
Jaroslaw Gawin <jaroslawx.gawin@intel.com> Signed-off-by:
Aleksandr Loktionov <aleksandr.loktionov@intel.com> Signed-off-by:
Arkadiusz Kubalewski <arkadiusz.kubalewski@intel.com> Tested-by:
Andrew Bowers <andrewx.bowers@intel.com> Signed-off-by:
-
- Sep 29, 2020
-
-
Sebastian Andrzej Siewior authored
in_interrupt() is ill defined and does not provide what the name suggests. The usage especially in driver code is deprecated and a tree wide effort to clean up and consolidate the (ab)usage of in_interrupt() and related checks is happening. In this case the checks cover only parts of the contexts in which these functions cannot be called. They fail to detect preemption or interrupt disabled invocations. As the functions which are invoked from the various places contain already a broad variety of checks (always enabled or debug option dependent) cover all invalid conditions already, there is no point in having inconsistent warnings in those drivers. Just remove them. Signed-off-by:
Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by:
Thomas Gleixner <tglx@linutronix.de> Reviewed-by:
Alexander Duyck <alexander.h.duyck@linux.intel.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
- Sep 28, 2020
-
-
Jakub Kicinski authored
Make use of the "shared port table" to convert i40e to the new infra. i40e did not have any reference tracking, locking is also dodgy because rtnl gets released while talking to FW, so port may get removed from the table while it's getting added etc. On the good side i40e does not seem to be using the ports for TX so we can remove the table from the driver state completely. Signed-off-by:
-
- Sep 25, 2020
-
-
Jesse Brandeburg authored
This takes care of all of the trivial W=1 fixes in the Intel Ethernet drivers, which allows developers and maintainers to build more of the networking tree with more complete warning checks. There are three classes of kdoc warnings fixed: - cannot understand function prototype: 'x' - Excess function parameter 'x' description in 'y' - Function parameter or member 'x' not described in 'y' All of the changes were trivial comment updates on function headers. Inspired by Lee Jones' series of wireless work to do the same. Compile tested only, and passes simple test of $ git ls-files *.[ch] | egrep drivers/net/ethernet/intel | \ xargs scripts/kernel-doc -none Signed-off-by:
Jesse Brandeburg <jesse.brandeburg@intel.com> Tested-by:
-
- Sep 14, 2020
-
-
Björn Töpel authored
Instead of checking in each iteration of the Rx packet processing loop, move the allocation out of the loop and do it once for each napi activation. For AF_XDP the rx_drop benchmark was improved by 6%. Signed-off-by:
-
Björn Töpel authored
The i40e NIC supports two flavors of HW descriptors, 16 and 32 byte. The latter has, obviously, room for more offloading information. However, the only fields of the 32B HW descriptor that is being used by the driver, is also available in the 16B descriptor. In other words; Reading and writing 32 bytes instead of 16 byte is a waste of bus bandwidth. This commit starts using 16 byte descriptors instead of 32 byte descriptors. For AF_XDP the rx_drop benchmark was improved by 2%. Signed-off-by:
-
Björn Töpel authored
The software prefetching of HW descriptors has a negative impact on the performance. Therefore, it is now removed. Performance for the rx_drop benchmark increased with 2%. Signed-off-by:
-
Li RongQing authored
refcount of rx_buffer page will be added here originally, so prefetchw is needed, but after commit 1793668c ("i40e/i40evf: Update code to better handle incrementing page count"), and refcount is not added every time, so change prefetchw as prefetch. Now it mainly services page_address(), but which accesses struct page only when WANT_PAGE_VIRTUAL or HASHED_PAGE_VIRTUAL is defined otherwise it returns address based on offset, so we prefetch it conditionally. Jakub suggested to define prefetch_page_address in a common header. Reported-by:
kernel test robot <lkp@intel.com> Suggested-by:
-
Li RongQing authored
After commit 759dc4a7 ("i40e: initialize our affinity_mask based on cpu_possible_mask"), NAPI IRQ affinity_mask is bind to all possible CPUs, not a fixed CPU Signed-off-by:
-
- Sep 09, 2020
-
-
Stefan Assmann authored
The for loop in i40e_set_vsi_promisc() reports errors via dev_err() but does not propagate the error up the call chain. Instead it continues the loop and potentially overwrites the reported error value. This results in the error being recorded in the log buffer, but the caller might never know anything went the wrong way. To avoid this situation i40e_set_vsi_promisc() needs to temporarily store the error after reporting it. This is still not optimal as multiple different errors may occur, so store the first error and hope that's the main issue. Fixes: 37d318d7 (i40e: Remove scheduling while atomic possibility) Reported-by:
Michal Schmidt <mschmidt@redhat.com> Signed-off-by:
-
Stefan Assmann authored
drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.c: In function ‘i40e_set_vsi_promisc’: drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.c:1176:14: error: ‘aq_ret’ may be used uninitialized in this function [-Werror=maybe-uninitialized] i40e_status aq_ret; In case the code inside the if statement and the for loop does not get executed aq_ret will be uninitialized when the variable gets returned at the end of the function. Avoid this by changing num_vlans from int to u16, so aq_ret always gets set. Making this change in additional places as num_vlans should never be negative. Fixes: 37d318d7 ("i40e: Remove scheduling while atomic possibility") Signed-off-by:
-
- Aug 31, 2020
-
-
Magnus Karlsson authored
Test for dma_need_sync earlier to increase performance. xsk_buff_dma_sync_for_cpu() takes an xdp_buff as parameter and from that the xsk_buff_pool reference is dug out. Perf shows that this dereference causes a lot of cache misses. But as the buffer pool is now sent down to the driver at zero-copy initialization time, we might as well use this pointer directly, instead of going via the xsk_buff and we can do so already in xsk_buff_dma_sync_for_cpu() instead of in xp_dma_sync_for_cpu. This gets rid of these cache misses. Throughput increases with 3% for the xdpsock l2fwd sample application on my machine. Signed-off-by:
-
Magnus Karlsson authored
Rename the AF_XDP zero-copy driver interface functions to better reflect what they do after the replacement of umems with buffer pools in the previous commit. Mostly it is about replacing the umem name from the function names with xsk_buff and also have them take the a buffer pool pointer instead of a umem. The various ring functions have also been renamed in the process so that they have the same naming convention as the internal functions in xsk_queue.h. This so that it will be clearer what they do and also for consistency. Signed-off-by:
-
Magnus Karlsson authored
Replace the explicit umem reference passed to the driver in AF_XDP zero-copy mode with the buffer pool instead. This in preparation for extending the functionality of the zero-copy mode so that umems can be shared between queues on the same netdev and also between netdevs. In this commit, only an umem reference has been added to the buffer pool struct. But later commits will add other entities to it. These are going to be entities that are different between different queue ids and netdevs even though the umem is shared between them. Signed-off-by:
-
- Aug 26, 2020
-
-
Tariq Toukan authored
Many device drivers use the same prefetch code structure to deal with small L1 cacheline size. Take this code into a function and call it from the drivers. Suggested-by:
Jakub Kicinski <jakub.kicinski@netronome.com> Signed-off-by:
Tariq Toukan <tariqt@mellanox.com> Reviewed-by:
Saeed Mahameed <saeedm@mellanox.com> Signed-off-by:
-
- Aug 14, 2020
-
-
Grzegorz Szczurek authored
Fix the reason of crashing system by add waiting time to finish reset recovery process before starting remove driver procedure. Now VSI is releasing if VSI is not in reset recovery mode. Without this fix it was possible to start remove driver if other processing command need reset recovery procedure which resulted in null pointer dereference. VSI used by the ethtool process has been cleared by remove driver process. [ 6731.508665] BUG: kernel NULL pointer dereference, address: 0000000000000000 [ 6731.508668] #PF: supervisor read access in kernel mode [ 6731.508670] #PF: error_code(0x0000) - not-present page [ 6731.508671] PGD 0 P4D 0 [ 6731.508674] Oops: 0000 [#1] SMP PTI [ 6731.508679] Hardware name: Intel Corporation S2600WT2R/S2600WT2R, BIOS SE5C610.86B.01.01.0021.032120170601 03/21/2017 [ 6731.508694] RIP: 0010:i40e_down+0x252/0x310 [i40e] [ 6731.508696] Code: c7 78 de fa c0 e8 61 02 3a c1 66 83 bb f6 0c 00 00 00 0f 84 bf 00 00 00 45 31 e4 45 31 ff eb 03 41 89 c7 48 8b 83 98 0c 00 00 <4a> 8b 3c 20 e8 a5 79 02 00 48 83 bb d0 0c 00 00 00 74 10 48 8b 83 [ 6731.508698] RSP: 0018:ffffb75ac7b3faf0 EFLAGS: 00010246 [ 6731.508700] RAX: 0000000000000000 RBX: ffff9c9874bd5000 RCX: 0000000000000007 [ 6731.508701] RDX: 0000000000000000 RSI: 0000000000000096 RDI: ffff9c987f4d9780 [ 6731.508703] RBP: ffffb75ac7b3fb30 R08: 0000000000005b60 R09: 0000000000000004 [ 6731.508704] R10: ffffb75ac64fbd90 R11: 0000000000000001 R12: 0000000000000000 [ 6731.508706] R13: ffff9c97a08e0000 R14: ffff9c97a08e0a68 R15: 0000000000000000 [ 6731.508708] FS: 00007f2617cd2740(0000) GS:ffff9c987f4c0000(0000) knlGS:0000000000000000 [ 6731.508710] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 6731.508711] CR2: 0000000000000000 CR3: 0000001e765c4006 CR4: 00000000003606e0 [ 6731.508713] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 [ 6731.508714] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 [ 6731.508715] Call Trace: [ 6731.508734] i40e_vsi_close+0x84/0x90 [i40e] [ 6731.508742] i40e_quiesce_vsi.part.98+0x3c/0x40 [i40e] [ 6731.508749] i40e_pf_quiesce_all_vsi+0x55/0x60 [i40e] [ 6731.508757] i40e_prep_for_reset+0x59/0x130 [i40e] [ 6731.508765] i40e_reconfig_rss_queues+0x5a/0x120 [i40e] [ 6731.508774] i40e_set_channels+0xda/0x170 [i40e] [ 6731.508778] ethtool_set_channels+0xe9/0x150 [ 6731.508781] dev_ethtool+0x1b94/0x2920 [ 6731.508805] dev_ioctl+0xc2/0x590 [ 6731.508811] sock_do_ioctl+0xae/0x150 [ 6731.508813] sock_ioctl+0x34f/0x3c0 [ 6731.508821] ksys_ioctl+0x98/0xb0 [ 6731.508828] __x64_sys_ioctl+0x1a/0x20 [ 6731.508831] do_syscall_64+0x57/0x1c0 [ 6731.508835] entry_SYSCALL_64_after_hwframe+0x44/0xa9 Fixes: 4b816446 ("i40e: Add common function for finding VSI by type") Signed-off-by:
Grzegorz Szczurek <grzegorzx.szczurek@intel.com> Signed-off-by:
-
Przemyslaw Patynowski authored
Trusted VF with unicast promiscuous mode set, could listen to TX traffic of other VFs. Set unicast promiscuous mode to RX traffic, if VSI has port VLAN configured. Rename misleading I40E_AQC_SET_VSI_PROMISC_TX bit to I40E_AQC_SET_VSI_PROMISC_RX_ONLY. Aligned unicast promiscuous with VLAN to the one without VLAN. Fixes: 6c41a760 ("i40e: Add promiscuous on VLAN support") Fixes: 3b120089 ("i40e: When in promisc mode apply promisc mode to Tx Traffic as well") Signed-off-by:
Przemyslaw Patynowski <przemyslawx.patynowski@intel.com> Signed-off-by:
-
- Jul 26, 2020
-
-
Andrii Nakryiko authored
Now that BPF program/link management is centralized in generic net_device code, kernel code never queries program id from drivers, so XDP_QUERY_PROG/XDP_QUERY_PROG_HW commands are unnecessary. This patch removes all the implementations of those commands in kernel, along the xdp_attachment_query(). This patch was compile-tested on allyesconfig. Signed-off-by:
Andrii Nakryiko <andriin@fb.com> Signed-off-by:
Alexei Starovoitov <ast@kernel.org> Link: https://lore.kernel.org/bpf/20200722064603.3350758-10-andriin@fb.com
-
- Jul 07, 2020
-
-
Luc Van Oostenryck authored
The method struct pci_error_handlers.error_detected() is defined and documented as taking an 'enum pci_channel_state' for the second argument, but most drivers use 'pci_channel_state_t' instead. This 'pci_channel_state_t' is not a typedef for the enum but a typedef for a bitwise type in order to have better/stricter typechecking. Consolidate everything by using 'pci_channel_state_t' in the method's definition, in the related helpers and in the drivers. Enforce use of 'pci_channel_state_t' by replacing 'enum pci_channel_state' with an anonymous 'enum'. Note: Currently, from a typechecking point of view this patch changes nothing because only the constants defined by the enum are bitwise, not the enum itself (sparse doesn't have the notion of 'bitwise enum'). This may change in some not too far future, hence the patch. [bhelgaas: squash in https://lore.kernel.org/r/20200702162651.49526-3-luc.vanoostenryck@gmail.com https://lore.kernel.org/r/20200702162651.49526-4-luc.vanoostenryck@gmail.com] Link: https://lore.kernel.org/r/20200702162651.49526-2-luc.vanoostenryck@gmail.com Signed-off-by:
Luc Van Oostenryck <luc.vanoostenryck@gmail.com> Signed-off-by:
Bjorn Helgaas <bhelgaas@google.com>
-
- Jul 01, 2020
-
-
Ciara Loftus authored
Interfaces already exist for dumping Rx and Tx descriptor information. Introduce another for doing the same for XDP descriptors. Signed-off-by:
Ciara Loftus <ciara.loftus@intel.com> Tested-by:
-
Ciara Loftus authored
Prior to this, only the Rx and Tx ring statistics were dumped. The XDP ring statistics are now dumped as well. Signed-off-by:
-
Ciara Loftus authored
Prior to this, only Rx and Tx ring statistics were accounted for. Signed-off-by:
-
Magnus Karlsson authored
Move the check if the HW Tx ring is full to outside the send loop. Currently it is checked for every single descriptor that we send. Instead, tell the send loop to only process a maximum number of packets equal to the number of available slots in the Tx ring. This way, we can remove the check inside the send loop to and gain some performance. Suggested-by:
Sridhar Samudrala <sridhar.samudrala@intel.com> Signed-off-by:
-
Magnus Karlsson authored
Eliminate a division in the napi_poll data path. This division is executed even though it is only needed in the rare case when there are not enough interrupt lines so they have to be shared between queue pairs. Instead, just test for this case and only execute the division if needed. The code has been lifted from the ice driver. Signed-off-by:
-
Magnus Karlsson authored
Improve the performance of the AF_XDP zero-copy Tx completion path. When there are no XDP buffers being sent using XDP_TX or XDP_REDIRECT, we do not have go through the SW ring to clean up any entries since the AF_XDP path does not use these. In these cases, just fast forward the next-to-use counter and skip going through the SW ring. The limit on the maximum number of entries to complete is also removed since the algorithm is now O(1). To simplify the code path, the maximum number of entries to complete for the XDP path is therefore also increased from 256 to 512 (the default number of Tx HW descriptors). This should be fine since the completion in the XDP path is faster than in the SKB path that has 256 as the maximum number. This patch provides around 4% throughput improvement for the l2fwd application in xdpsock on my machine. Signed-off-by:
-
Arkadiusz Kubalewski authored
After OS requests to down a link on a physical network port, the traffic is no longer being processed but the physical link with a link partner is still established. Currently there is a feature (Link down on close) which allows to physically bring the link down (after OS request). With this patch new feature with similar capability is introduced: TOTAL_PORT_SHUTDOWN Allows to physically disable the link on the NIC's port. If enabled, (after link down request from the OS) no link, traffic or led activity is possible on that port. If I40E_FLAG_TOTAL_PORT_SHUTDOWN is enabled, the I40E_FLAG_LINK_DOWN_ON_CLOSE_ENABLED must be explicitly forced to true and cannot be disabled at that time. The functionalities are exclusive in terms of configuration, but they also have similar behavior (allowing to disable physical link of the port), with following differences: - LINK_DOWN_ON_CLOSE_ENABLED is configurable at host OS run-time and is supported by whole family of 7xx Intel Ethernet Controllers - TOTAL_PORT_SHUTDOWN may be enabled only before OS loads (in BIOS) only if motherboard's BIOS and NIC's FW has support of it - when LINK_DOWN_ON_CLOSE_ENABLED is used, the link is being brought down by sending phy_type=0 to NIC's FW - when TOTAL_PORT_SHUTDOWN is used, phy_type is not altered, instead the link is being brought down by clearing bit (I40E_AQ_PHY_ENABLE_LINK) in abilities field of i40e_aq_set_phy_config structure Introduced changes: - new private flag I40E_FLAG_TOTAL_PORT_SHUTDOWN for handling the feature - probe of NVM if the feature was enabled at driver's port initialization - special handling on link-down procedure to let FW physically shutdown the port if the feature was enabled Signed-off-by:
-
Jeff Kirsher authored
Convert all the remaining 'fall through" code comments to the newer 'fallthrough;' keyword. Suggested-by:
Joe Perches <joe@perches.com> Signed-off-by:
Jeff Kirsher <jeffrey.t.kirsher@intel.com> Tested-by:
-
- Jun 26, 2020
-
-
Aleksandr Loktionov authored
In some occasions task held spinlock (mac_filter_hash_lock), while being rescheduled due to admin queue mutex_lock. The struct i40e_spinlock asq_spinlock, which later expands to struct mutex spinlock. Moved i40e_aq_set_vsi_multicast_promiscuous(), i40e_aq_set_vsi_unicast_promiscuous(), i40e_aq_set_vsi_mc_promisc_on_vlan(), and i40e_aq_set_vsi_uc_promisc_on_vlan() outside of atomic context. Without this patch there is a race condition, which might result in scheduling while in atomic context. The race condition is between the thread, which holds mac_filter_hash_lock, while trying to acquire an admin queue mutex and a thread, which already has said admin queue mutex. The thread, which holds spinlock, fails to acquire the mutex, which causes this thread to sleep. Signed-off-by:
-
