- 17 May, 2020 10 commits
-
-
John Hubbard authored
This code was using get_user_pages_fast(), in a "Case 2" scenario (DMA/RDMA), using the categorization from [1]. That means that it's time to convert the get_user_pages_fast() + put_page() calls to pin_user_pages_fast() + unpin_user_pages() calls. There is some helpful background in [2]: basically, this is a small part of fixing a long-standing disconnect between pinning pages, and file systems' use of those pages. [1] Documentation/core-api/pin_user_pages.rst [2] "Explicit pinning of user-space pages": https://lwn.net/Articles/807108/ Cc: David S. Miller <davem@davemloft.net> Cc: Jakub Kicinski <kuba@kernel.org> Cc: netdev@vger.kernel.org Cc: linux-rdma@vger.kernel.org Cc: rds-devel@oss.oracle.com Signed-off-by:John Hubbard <jhubbard@nvidia.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
David S. Miller authored
Florian Westphal says: ==================== mptcp: do not block on subflow socket This series reworks mptcp_sendmsg logic to avoid blocking on the subflow socket. It does so by removing the wait loop from mptcp_sendmsg_frag helper. In order to do that, it moves prerequisites that are currently handled in mptcp_sendmsg_frag (and cause it to wait until they are met, e.g. frag cache refill) into the callers. The worker can just reschedule in case no subflow socket is ready, since it can't wait -- doing so would block other work items and doesn't make sense anyway because we should not (re)send data in case resources are already low. The sendmsg path can use the existing wait logic until memory becomes available. Because large send requests can result in multiple mptcp_sendmsg_frag calls from mptcp_sendmsg, we may need to restart the socket lookup in case subflow can't accept more data or memory is low. Doing so blocks on the mptcp socket, and existing wait handling releases the msk lock while blocking. Lastly, no need to use GFP_ATOMIC for extension allocation: extend __skb_ext_alloc with gfp_t arg instead of hard-coded ATOMIC and then relax the allocation constraints for mptcp case: those requests occur in process context. ==================== Signed-off-by: -
Florian Westphal authored
mptcp calls this from the transmit side, from process context. Allow a sleeping allocation instead of unconditional GFP_ATOMIC. Acked-by:
Paolo Abeni <pabeni@redhat.com> Signed-off-by:
Florian Westphal <fw@strlen.de> Signed-off-by:
-
Florian Westphal authored
previous patches made sure we only call into this function when these prerequisites are met, so no need to wait on the subflow socket anymore. Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/7Acked-by:
-
Florian Westphal authored
The mptcp_sendmsg_frag helper contains a loop that will wait on the subflow sk. It seems preferrable to only wait in mptcp_sendmsg() when blocking io is requested. mptcp_sendmsg already has such a wait loop that is used when no subflow socket is available for transmission. This is another preparation patch that makes sure we call mptcp_sendmsg_frag only if the page frag cache has been refilled. Followup patch will remove the wait loop from mptcp_sendmsg_frag(). The retransmit worker doesn't need to do this refill as it won't transmit new mptcp-level data. Acked-by:
-
Florian Westphal authored
The mptcp_sendmsg_frag helper contains a loop that will wait on the subflow sk. It seems preferrable to only wait in mptcp_sendmsg() when blocking io is requested. mptcp_sendmsg already has such a wait loop that is used when no subflow socket is available for transmission. This is a preparation patch that makes sure we call mptcp_sendmsg_frag only if a skb extension has been allocated. Moreover, such allocation currently uses GFP_ATOMIC while it could use sleeping allocation instead. Followup patches will remove the wait loop from mptcp_sendmsg_frag() and will allow to do a sleeping allocation for the extension. Acked-by:
-
Florian Westphal authored
The transmit loop continues to xmit new data until an error is returned or all data was transmitted. For the blocking i/o case, this means that tcp_sendpages() may block on the subflow until more space becomes available, i.e. we end up sleeping with the mptcp socket lock held. Instead we should check if a different subflow is ready to be used. This restarts the subflow sk lookup when the tx operation succeeded and the tcp subflow can't accept more data or if tcp_sendpages indicates -EAGAIN on a blocking mptcp socket. In that case we also need to set the NOSPACE bit to make sure we get notified once memory becomes available. In case all subflows are busy, the existing logic will wait until a subflow is ready, releasing the mptcp socket lock while doing so. The mptcp worker already sets DONTWAIT, so no need to make changes there. v2: * set NOSPACE bit * add a comment to clarify that mptcp-sk sndbuf limits need to be checked as well. Signed-off-by:
-
Florian Westphal authored
Its not enough to check for available tcp send space. We also hold on to transmitted data for mptcp-level retransmits. Right now we will send more and more data if the peer can ack data at the tcp level fast enough, since that frees up tcp send buffer space. But we also need to check that data was acked and reclaimed at the mptcp level. Therefore add needed check in mptcp_sendmsg, flush tcp data and wait until more mptcp snd space becomes available if we are over the limit. Before we wait for more data, also make sure we start the retransmit timer if we ran out of sndbuf space. Otherwise there is a very small chance that we wait forever: * receiver is waiting for data * sender is blocked because mptcp socket buffer is full * at tcp level, all data was acked * mptcp-level snd_una was not updated, because last ack that acknowledged the last data packet carried an older MPTCP-ack. Restarting the retransmit timer avoids this problem: if TCP subflow is idle, data is retransmitted from the RTX queue. New data will make the peer send a new, updated MPTCP-Ack. Signed-off-by:
-
Florian Westphal authored
Paolo noticed that ssk_check_wmem() has same pattern, so add/use common helper for both places. Suggested-by:
-
David Ahern authored
The 'pref medium' attribute was moved in iproute2 to be near the prefix which is where it applies versus after the last nexthop. The nexthop tests were updated to drop the string from route checking, but it crept in again with the compat tests. Fixes: 4dddb5be ("selftests: net: add new testcases for nexthop API compat mode sysctl") Signed-off-by:
David Ahern <dsahern@gmail.com> Cc: Roopa Prabhu <roopa@cumulusnetworks.com> Signed-off-by:
-
- 16 May, 2020 19 commits
-
-
David S. Miller authored
Alex Elder says: ==================== net: ipa: sc7180 suspend/resume This series permits suspend/resume to work for the IPA driver on the Qualcomm SC7180 SoC. The IPA version on this SoC requires interrupts to be enabled when the suspend and resume callbacks are made, and the first patch moves away from using the noirq variants. The second patch fixes a problem with resume that occurs because pending interrupts were being cleared before starting a channel. ==================== Signed-off-by: -
Alex Elder authored
In gsi_channel_start() there is harmless-looking comment "Clear the channel's event ring interrupt in case it's pending". The intent was to avoid getting spurious interrupts when first bringing up a channel. However we now use channel stop/start to implement suspend and resume, and an interrupt pending at the time we resume is actually something we don't want to ignore. The very first time we bring up the channel we do not expect an interrupt to be pending, and even if it were, the effect would simply be to schedule NAPI on that channel, which would find nothing to do, which is not a problem. Stop clearing any pending IEOB interrupt in gsi_channel_start(). That leaves one caller of the trivial function gsi_isr_ieob_clear(). Get rid of that function and just open-code it in gsi_isr_ieob() instead. This fixes a problem where suspend/resume IPA v4.2 would get stuck when resuming after a suspend. Signed-off-by:
Alex Elder <elder@linaro.org> Signed-off-by:
-
Alex Elder authored
Use the suspend and resume callbacks rather than suspend_noirq and resume_noirq. With IPA v4.2, we use the CHANNEL_STOP command to implement a suspend, and without interrupts enabled, that command won't complete. Signed-off-by:
-
David S. Miller authored
Ido Schimmel says: ==================== mlxsw: Reorganize trap data This patch set does not include any functional changes. It merely reworks the internal storage of traps, trap groups and trap policers in mlxsw to each use a single array. These changes allow us to get rid of the multiple arrays we currently have for traps, which make the trap data easier to validate and extend with more per-trap information in the future. It will also allow us to more easily add per-ASIC traps in future submissions. Last two patches include minor changes to devlink-trap selftests. Tested with existing devlink-trap selftests. ==================== Signed-off-by: -
Ido Schimmel authored
It can be derived dynamically from the trap's name, so drop it. Signed-off-by:
Ido Schimmel <idosch@mellanox.com> Reviewed-by:
Jiri Pirko <jiri@mellanox.com> Signed-off-by:
-
Ido Schimmel authored
One blank line is enough. Signed-off-by:
-
Ido Schimmel authored
Each trap registered with devlink is mapped to one or more Rx listeners. These listeners allow the switch driver (e.g., mlxsw_spectrum) to register a function that is called when a packet is received (trapped) for a specific reason. Currently, three arrays are used to describe the mapping between the logical devlink traps and the Rx listeners. Instead, get rid of these arrays and store all the information in one array that is easier to validate and extend with more per-trap information. Signed-off-by:
-
Ido Schimmel authored
Use one array to store all the information about all the trap groups instead of hard coding it in code. This will be used in future patches to disable certain functionality (e.g., policer binding) on a trap group basis. Signed-off-by:
-
Ido Schimmel authored
Instead of maintaining an array of policers and a linked list, only maintain an array. Signed-off-by:
-
Ido Schimmel authored
'struct mlxsw_sp_trap_policer_item' is only used in one file, so move it there. Signed-off-by:
-
Heiner Kallweit authored
After having switched to devm_mdiobus_register() also this remaining call to mdiobus_unregister() can be removed. Signed-off-by:
Heiner Kallweit <hkallweit1@gmail.com> Signed-off-by:
-
David S. Miller authored
Jakub Kicinski says: ==================== ethtool: set_channels: add a few more checks There seems to be a few more things we can check in the core before we call drivers' ethtool_ops->set_channels. Adding the checks to the core simplifies the drivers. This set only includes changes to the NFP driver as an example. There is a small risk in the first patch that someone actually purposefully accepts a strange configuration without RX or TX channels, but I couldn't find such a driver in the tree. ==================== Signed-off-by: -
Jakub Kicinski authored
Don't call drivers if nothing changed. Netlink code already contains this logic. Signed-off-by:
Jakub Kicinski <kuba@kernel.org> Reviewed-by:
Michal Kubecek <mkubecek@suse.cz> Signed-off-by:
-
Jakub Kicinski authored
Core will now perform this check. Signed-off-by:
-
Jakub Kicinski authored
Having a channel config with no ability to RX or TX traffic is clearly wrong. Check for this in the core so the drivers don't have to. Signed-off-by:
-
Christoph Paasch authored
RFC8684 allows to send 32-bit DATA_ACKs as long as the peer is not sending 64-bit data-sequence numbers. The 64-bit DSN is only there for extreme scenarios when a very high throughput subflow is combined with a long-RTT subflow such that the high-throughput subflow wraps around the 32-bit sequence number space within an RTT of the high-RTT subflow. It is thus a rare scenario and we should try to use the 32-bit DATA_ACK instead as long as possible. It allows to reduce the TCP-option overhead by 4 bytes, thus makes space for an additional SACK-block. It also makes tcpdumps much easier to read when the DSN and DATA_ACK are both either 32 or 64-bit. Signed-off-by:
Christoph Paasch <cpaasch@apple.com> Reviewed-by:
Matthieu Baerts <matthieu.baerts@tessares.net> Signed-off-by:
-
Nicolas Dichtel authored
The goal is to be able to inherit the initial devconf parameters from the current netns, ie the netns where this new netns has been created. This is useful in a containers environment where /proc/sys is read only. For example, if a pod is created with specifics devconf parameters and has the capability to create netns, the user expects to get the same parameters than his 'init_net', which is not the real init_net in this case. Signed-off-by:
Nicolas Dichtel <nicolas.dichtel@6wind.com> Signed-off-by:
-
Ioana Ciornei authored
Add driver level bulking to the XDP_TX action. An array of frame descriptors is held for each Tx frame queue and populated accordingly when the action returned by the XDP program is XDP_TX. The frames will be actually enqueued only when the array is filled. At the end of the NAPI cycle a flush on the queued frames is performed in order to enqueue the remaining FDs. Signed-off-by:
Ioana Ciornei <ioana.ciornei@nxp.com> Signed-off-by:
-
Kevin Lo authored
This patch makes checkpatch happy for tabs Signed-off-by:
Kevin Lo <kevlo@kevlo.org> Reviewed-by:
Florian Fainelli <f.fainelli@gmail.com> Signed-off-by:
-
- 15 May, 2020 11 commits
-
-
git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linuxDavid S. Miller authored
Saeed Mahameed says: ==================== mlx5-updates-2020-05-15 mlx5 core and mlx5e (netdev) updates: 1) Two fixes for release all FW pages support. 2) Improvement in calculating the send queue stop room on tx 3) Flow steering auto-groups creation improvements 4) TC offload fix for Connection tracking with NAT action 5) IPoIB support for self looback to allow communication between ipoib pkey child interfaces on the same host. 6) DCBNL cleanup to avoid #ifdef DCBNL all over the main mlx5e code 7) Small and trivial code cleanup ==================== Signed-off-by: -
Nathan Chancellor authored
When building with Clang: In file included from drivers/net/ethernet/ti/am65-cpsw-ethtool.c:15: drivers/net/ethernet/ti/am65-cpts.h:58:12: warning: unused function 'am65_cpts_ns_gettime' [-Wunused-function] static s64 am65_cpts_ns_gettime(struct am65_cpts *cpts) ^ drivers/net/ethernet/ti/am65-cpts.h:63:12: warning: unused function 'am65_cpts_estf_enable' [-Wunused-function] static int am65_cpts_estf_enable(struct am65_cpts *cpts, ^ drivers/net/ethernet/ti/am65-cpts.h:69:13: warning: unused function 'am65_cpts_estf_disable' [-Wunused-function] static void am65_cpts_estf_disable(struct am65_cpts *cpts, int idx) ^ 3 warnings generated. These functions need to be marked as inline, which adds __maybe_unused, to avoid these warnings, which is the pattern for stub functions. Fixes: ec008fa2 ("ethernet: ti: am65-cpts: add routines to support taprio offload") Link: https://github.com/ClangBuiltLinux/linux/issues/1026Signed-off-by:Nathan Chancellor <natechancellor@gmail.com> Signed-off-by:
-
Tariq Toukan authored
Take DCBNL-related definitions out of the common en.h header, Use a dedicated header file for exposing them. Some need not to be exposed, use them locally in the .c file. Use stubs to eliminate use of CONFIG_MLX5_CORE_EN_DCB in the generic control flows. Signed-off-by:
Tariq Toukan <tariqt@mellanox.com> Signed-off-by:
Saeed Mahameed <saeedm@mellanox.com>
-
Maxim Mikityanskiy authored
Currently, different formulas are used to estimate the space that may be taken by WQEs in the SQ during a single packet transmit. This space is called stop room, and it's checked in the end of packet transmit to find out if the next packet could overflow the SQ. If it could, the driver tells the kernel to stop sending next packets. Many factors affect the stop room: 1. Padding with NOPs to avoid WQEs spanning over page boundaries. 2. Enabled and disabled offloads (TLS, upcoming MPWQE). 3. The maximum size of a WQE. The padding is performed before every WQE if it doesn't fit the current page. The current formula assumes that only one padding will be required per packet, and it doesn't take into account that the WQEs posted during the transmission of a single packet might exceed the page size in very rare circumstances. For example, to hit this condition with 4096-byte pages, TLS offload will have to interrupt an almost-full MPWQE session, be in the resync flow and try to transmit a near to maximum amount of data. To avoid SQ overflows in such rare cases after MPWQE is added, this patch introduces a more robust formula to estimate the stop room. The new formula uses the fact that a WQE of size X will not require more than X-1 WQEBBs of padding. More exact estimations are possible, but they result in much more complex and error-prone code for little gain. Before this patch, the TLS stop room included space for both INNOVA and ConnectX TLS offloads that couldn't run at the same time anyway, so this patch accounts only for the active one. Signed-off-by:
Maxim Mikityanskiy <maximmi@mellanox.com> Reviewed-by:
-
Erez Shitrit authored
After enabled loopback packets for IPoIB, we need to drop these packets that this HCA has replicated and came back to the same interface that sent them. Fixes: 4c6c615e ("net/mlx5e: IPoIB, Add PKEY child interface nic profile") Signed-off-by:
Erez Shitrit <erezsh@mellanox.com> Reviewed-by:
Alex Vesker <valex@mellanox.com> Signed-off-by:
-
Erez Shitrit authored
Enable loopback of unicast and multicast traffic for IPoIB enhanced mode. This will allow interfaces with the same pkey to communicate between them e.g cloned interfaces that located in different namespaces. Signed-off-by:
-
Roi Dayan authored
It could be a chain of rules will do action CT again after CT NAT Before this fix matching will break as we get into the CT table after NAT changes and not CT NAT. Fix this by adding pre ct and pre ct nat tables to skip ct/ct_nat tables and go straight to post_ct table if ct/nat was already done. Signed-off-by:
Roi Dayan <roid@mellanox.com> Reviewed-by:
Paul Blakey <paulb@mellanox.com> Signed-off-by:
-
Eran Ben Elisha authored
Move mlx5_read_internal_timer() into lib/clock.c file as it is being used there. As such, make this function a static one. In addition, rearrange headers include to support function move. Signed-off-by:
Eran Ben Elisha <eranbe@mellanox.com> Reviewed-by:
Aya Levin <ayal@mellanox.com> Signed-off-by:
-
Paul Blakey authored
Currently, if one thread tries to add an entry to an autogrouped table with no free matching group, while another thread is in the process of creating a new matching autogroup, it doesn't wait for the new group creation, and creates an unnecessary new autogroup. Instead of skipping inactive, wait on the write lock of those groups. Signed-off-by:
Mark Bloch <markb@mellanox.com> Reviewed-by:
Maor Gottlieb <maorg@mellanox.com> Signed-off-by:
-
Parav Pandit authored
mlx5_unload_one() is done with cleanup = true only once. So instead of doing health wq drain inside the if(), directly do during PCI device removal. Signed-off-by:
Parav Pandit <parav@mellanox.com> Reviewed-by:
Moshe Shemesh <moshe@mellanox.com> Signed-off-by:
-
Parav Pandit authored
Having multiple error unwinding path are error prone. Lets have just one error unwinding path. Signed-off-by:
-
