- 01 May, 2019 40 commits
-
-
Vinicius Costa Gomes authored
Right now, this isn't a problem, but the next commit allows schedules to be added during runtime. When a new schedule transitions from the inactive to the active state ("admin" -> "oper") the previous one can be freed, if it's freed just after the RCU read lock is released, we may access an invalid entry. So, we should take care to protect the dequeue() flow, so all the places that access the entries are protected by the RCU read lock. Signed-off-by:Vinicius Costa Gomes <vinicius.gomes@intel.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
David S. Miller authored
Yuchung Cheng says: ==================== undo congestion window on spurious SYN or SYNACK timeout Linux TCP currently uses the initial congestion window of 1 packet if multiple SYN or SYNACK timeouts per RFC6298. However such timeouts are often spurious on wireless or cellular networks that experience high delay variances (e.g. ramping up dormant radios or local link retransmission). Another case is when the underlying path is longer than the default SYN timeout (e.g. 1 second). In these cases starting the transfer with a minimal congestion window is detrimental to the performance for short flows. One naive approach is to simply ignore SYN or SYNACK timeouts and always use a larger or default initial window. This approach however risks pouring gas to the fire when the network is already highly congested. This is particularly true in data center where application could start thousands to millions of connections over a single or multiple hosts resulting in high SYN drops (e.g. incast). This patch-set detects spurious SYN and SYNACK timeouts upon completing the handshake via the widely-supported TCP timestamp options. Upon such events the sender reverts to the default initial window to start the data transfer so it gets best of both worlds. This patch-set supports this feature for both active and passive as well as Fast Open or regular connections. ==================== Signed-off-by: -
Yuchung Cheng authored
Relocate the congestion window initialization from tcp_init_metrics() to tcp_init_transfer() to improve code readability. Signed-off-by:
Yuchung Cheng <ycheng@google.com> Signed-off-by:
Neal Cardwell <ncardwell@google.com> Signed-off-by:
Soheil Hassas Yeganeh <soheil@google.com> Signed-off-by:
Eric Dumazet <edumazet@google.com> Signed-off-by:
-
Yuchung Cheng authored
Use a helper to consolidate two identical code block for passive TFO. Signed-off-by:
-
Yuchung Cheng authored
This patch makes passive Fast Open reverts the cwnd to default initial cwnd (10 packets) if the SYNACK timeout is spurious. Passive Fast Open uses a full socket during handshake so it can use the existing undo logic to detect spurious retransmission by recording the first SYNACK timeout in key state variable retrans_stamp. Upon receiving the ACK of the SYNACK, if the socket has sent some data before the timeout, the spurious timeout is detected by tcp_try_undo_recovery() in tcp_process_loss() in tcp_ack(). But if the socket has not send any data yet, tcp_ack() does not execute the undo code since no data is acknowledged. The fix is to check such case explicitly after tcp_ack() during the ACK processing in SYN_RECV state. In addition this is checked in FIN_WAIT_1 state in case the server closes the socket before handshake completes. Signed-off-by:
-
Yuchung Cheng authored
TCP sender would use congestion window of 1 packet on the second SYN and SYNACK timeout except passive TCP Fast Open. This makes passive TFO too aggressive and unfair during congestion at handshake. This patch fixes this issue so TCP (fast open or not, passive or active) always conforms to the RFC6298. Note that tcp_enter_loss() is called only once during recurring timeouts. This is because during handshake, high_seq and snd_una are the same so tcp_enter_loss() would incorrect set the undo state variables multiple times. Signed-off-by:
-
Yuchung Cheng authored
Linux implements RFC6298 and use an initial congestion window of 1 upon establishing the connection if the SYNACK packet is retransmitted 2 or more times. In cellular networks SYNACK timeouts are often spurious if the wireless radio was dormant or idle. Also some network path is longer than the default SYNACK timeout. In both cases falsely starting with a minimal cwnd are detrimental to performance. This patch avoids doing so when the final ACK's TCP timestamp indicates the original SYNACK was delivered. It remembers the original SYNACK timestamp when SYNACK timeout has occurred and re-uses the function to detect spurious SYN timeout conveniently. Note that a server may receives multiple SYNs from and immediately retransmits SYNACKs without any SYNACK timeout. This often happens on when the client SYNs have timed out due to wireless delay above. In this case since the server will still use the default initial congestion (e.g. 10) because tp->undo_marker is reset in tcp_init_metrics(). This is an intentional design because packets are not lost but delayed. This patch only covers regular TCP passive open. Fast Open is supported in the next patch. Signed-off-by:
-
Yuchung Cheng authored
Detecting spurious SYNACK timeout using timestamp option requires recording the exact SYNACK skb timestamp. Previously the SYNACK sent timestamp was stamped slightly earlier before the skb was transmitted. This patch uses the SYNACK skb transmission timestamp directly. Signed-off-by:
-
Yuchung Cheng authored
Linux implements RFC6298 and use an initial congestion window of 1 upon establishing the connection if the SYN packet is retransmitted 2 or more times. In cellular networks SYN timeouts are often spurious if the wireless radio was dormant or idle. Also some network path is longer than the default SYN timeout. Having a minimal cwnd on both cases are detrimental to TCP startup performance. This patch extends TCP undo feature (RFC3522 aka TCP Eifel) to detect spurious SYN timeout via TCP timestamps. Since tp->retrans_stamp records the initial SYN timestamp instead of first retransmission, we have to implement a different undo code additionally. The detection also must happen before tcp_ack() as retrans_stamp is reset when SYN is acknowledged. Note this patch covers both active regular and fast open. Signed-off-by:
-
Yuchung Cheng authored
Previously if an active TCP open has SYN timeout, it always undo the cwnd upon receiving the SYNACK. This is because tcp_clean_rtx_queue would reset tp->retrans_stamp when SYN is acked, which fools then tcp_try_undo_loss and tcp_packet_delayed. Addressing this issue is required to properly support undo for spurious SYN timeout. Fixing this is tricky -- for active TCP open tp->retrans_stamp records the time when the handshake starts, not the first retransmission time as the name may suggest. The simplest fix is for tcp_packet_delayed to ensure it is valid before comparing with other timestamp. One side effect of this change is active TCP Fast Open that incurred SYN timeout. Upon receiving a SYN-ACK that only acknowledged the SYN, it would immediately retransmit unacknowledged data in tcp_ack() because the data is marked lost after SYN timeout. But the retransmission would have an incorrect ack sequence number since rcv_nxt has not been updated yet tcp_rcv_synsent_state_process(), the retransmission needs to properly handed by tcp_rcv_fastopen_synack() like before. Signed-off-by:
-
Gustavo A. R. Silva authored
Replace "pass through" with a proper "fall through" annotation in order to fix the following warning: drivers/net/netdevsim/bus.c: In function ‘new_device_store’: drivers/net/netdevsim/bus.c:170:14: warning: this statement may fall through [-Wimplicit-fallthrough=] port_count = 1; ~~~~~~~~~~~^~~ drivers/net/netdevsim/bus.c:172:2: note: here case 2: ^~~~ Warning level 3 was used: -Wimplicit-fallthrough=3 This fix is part of the ongoing efforts to enable -Wimplicit-fallthrough Signed-off-by:
Gustavo A. R. Silva <gustavo@embeddedor.com> Acked-by:
Jakub Kicinski <jakub.kicinski@netronome.com> Signed-off-by:
-
Gustavo A. R. Silva authored
In preparation to enabling -Wimplicit-fallthrough, mark switch cases where we are expecting to fall through. This patch fixes the following warning: drivers/net/ethernet/sfc/mcdi_port.c: In function ‘efx_mcdi_phy_decode_link’: ./include/linux/compiler.h:77:22: warning: this statement may fall through [-Wimplicit-fallthrough=] # define unlikely(x) __builtin_expect(!!(x), 0) ^~~~~~~~~~~~~~~~~~~~~~~~~~ ./include/asm-generic/bug.h:125:2: note: in expansion of macro ‘unlikely’ unlikely(__ret_warn_on); \ ^~~~~~~~ drivers/net/ethernet/sfc/mcdi_port.c:344:3: note: in expansion of macro ‘WARN_ON’ WARN_ON(1); ^~~~~~~ drivers/net/ethernet/sfc/mcdi_port.c:345:2: note: here case MC_CMD_FCNTL_OFF: ^~~~ Warning level 3 was used: -Wimplicit-fallthrough=3 This patch is part of the ongoing efforts to enable -Wimplicit-fallthrough. Signed-off-by:Edward Cree <ecree@solarflare.com> Signed-off-by:
-
Moshe Shemesh authored
The devlink health reporters create/destroy and user commands currently use the devlink->lock as a locking mechanism. Different reporters have different rules in the driver and are being created/destroyed during different stages of driver load/unload/running. So during execution of a reporter recover the flow can go through another reporter's destroy and create. Such flow leads to deadlock trying to lock a mutex already held. With the new locking mechanism the different reporters share mutex lock only to protect access to shared reporters list. Added refcount per reporter, to protect the reporters from destroy while being used. Signed-off-by:
Moshe Shemesh <moshe@mellanox.com> Signed-off-by:
Jiri Pirko <jiri@mellanox.com> Acked-by:
-
David S. Miller authored
Igor Russkikh says: ==================== net: atlantic: Aquantia driver updates 2019-04 This patchset contains various improvements: - Work targeting link up speedups: link interrupt introduced, some other logic changes to imrove this. - FW operations securing with mutex - Counters and statistics logic improved by Dmitry - read out of chip temperature via hwmon interface implemented by Yana and Nikita. v4 changes: - remove drvinfo_exit noop - 64bit stats should be readed out sequentially (lsw, then msw) declare 64bit read ops for that v3 changes: - temp ops renamed to phy_temp ops - mutex commits squashed for better structure v2 changes: - use threaded irq for link state handling - rework hwmon via devm_hwmon_device_register_with_info Extra comments on review from Andrew: - direct device name pointer is used in hwmon registration. This causes hwmon device to derive possible interface name changes - Will consider sanity checks for firmware mutex lock separately. Right now there is no single point exsists where such check could be easily added. - There is no way now to fetch and configure min/max/crit temperatures via FW. Will investigate this separately. ==================== Signed-off-by: -
Nikita Danilov authored
Some device ids were never released and does not exist. Cleanup these. Signed-off-by:
Nikita Danilov <ndanilov@aquantia.com> Signed-off-by:
Igor Russkikh <igor.russkikh@aquantia.com> Signed-off-by:
-
Dmitry Bogdanov authored
DMA counters are 64 bit and we can fetch that to reduce counter overflow, espesially on byte counters. Tested-by:
Dmitry Bogdanov <dmitry.bogdanov@aquantia.com> Signed-off-by:
-
Dmitry Bogdanov authored
aq_nic_update_ndev_stats pushes statistics to ndev->stats from system interface. This is not always good because it counts packets/bytes before any of rx filters (including mac filter). Its better to report the packet/bytes statistics from DMA counters which gives actual values of data transferred over pci. System level stats is still available via ethtool. Signed-off-by:
-
Dmitry Bogdanov authored
This improves ethtool -S usage, where stats are now actual on each request. Before that stats only were updated at service timer period. Tested-by:
-
Igor Russkikh authored
Service timer callback fetches statistics from FW and that may cause a long delay in error cases. We also now need to use fw mutex to prevent concurrent access to FW, thus - extract that logic from timer callback into the job in the separate work queue. Signed-off-by:
-
Nikita Danilov authored
Some of FW operations could be invoked simultaneously, from f.e. ethtool context and from service service activity work. Here we introduce a fw mutex to secure and serialize access to FW logic. Signed-off-by:
-
Igor Russkikh authored
Typo in msi code. No much impact though. Signed-off-by:
-
Igor Russkikh authored
Improve for better readability Signed-off-by:
-
Igor Russkikh authored
Original code detected link only after 1 sec is passed after up. Here we replace this with direct service callback which updates link status immediately Signed-off-by:
-
Igor Russkikh authored
Here we define and request an extra interrupt line, assign it on link isr handler and restructure abit aq_pci code to better support that. We also remove logic for using different timer intervals depending on link state, since thats now useless. Signed-off-by:
-
Nikita Danilov authored
We need this to schedule link interrupt handling and various service tasks. Signed-off-by:
-
Igor Russkikh authored
Define link interrupt handler Signed-off-by:
-
Igor Russkikh authored
Declare macroes and nic fields to support link interrupt handling Signed-off-by:
-
Yana Esina authored
Added support for hwmon api to fetch out chip temperature Signed-off-by:
Yana Esina <yana.esina@aquantia.com> Signed-off-by:
Nikita Danilov <nikita.danilov@aquantia.com> Signed-off-by:
Andrew Lunn <andrew@lunn.ch> Signed-off-by:
-
Yana Esina authored
Ability to read the chip temperature from memory via hwmon interface Signed-off-by:
-
Heiner Kallweit authored
According to Neil who reported the issue leading to this workaround, the workaround is no longer needed since version 5.0. So let's remove it. This was the bug report leading to the workaround: https://bugzilla.kernel.org/show_bug.cgi?id=201081Signed-off-by:
Heiner Kallweit <hkallweit1@gmail.com> Tested-by:
Neil MacLeod <neil@nmacleod.com> Signed-off-by:
-
David S. Miller authored
Heiner Kallweit says: ==================== r8169: improve eri function handling This series aims at improving and simplifying the eri functions. No functional change intended. ==================== Signed-off-by: -
Heiner Kallweit authored
Fortunately in one place there's a comment explaining what toggling this bit does. So let's create a helper for it. Signed-off-by:
-
Heiner Kallweit authored
Add helpers rtl_eri_set_bits and rtl_eri_clear_bits to improve readability of the code. Signed-off-by:
-
Heiner Kallweit authored
In basically all eri function calls the type argument is ERIAR_EXGMAC. Therefore make it the default. Signed-off-by:
-
David S. Miller authored
Andrew Lunn says: ==================== Convert mv88e6060 to mdio device This patchset builds upon the previous patches to mv88e6060. It adds support for probing the switch as an MDIO device and then removes the legacy probe method. Since this is the last device supporting legacy probe, this allows legacy probe to be removed, originally planned to be removed in 4.17, but took a bit longer. This change to the mv88e6060 is more risky than the previous patchset. Some attempts to test it have been made, by hacking the driver to match on an mv88e6352 so that it probes. These changes are all about probe, so it is a reasonable test. But testing on a real mv88e6060 would be great. ==================== Signed-off-by: -
Andrew Lunn authored
Now that the code to support the legacy binding has been removed, remove the documentation for it. Signed-off-by:
-
Andrew Lunn authored
Now that all drivers can be probed using more traditional methods, remove the legacy probe code. Signed-off-by:
-
Andrew Lunn authored
Now that the driver can be probed as an mdio device, remove the legacy DSA platform device probing. Signed-off-by:
-
Andrew Lunn authored
Probing DSA devices as platform devices has been superseded by using normal bus drivers. Add support for probing the mv88e6060 device as an mdio device. Signed-off-by:
-
David S. Miller authored
Vladimir Oltean says: ==================== Improvements to DSA core VLAN manipulation In preparation of submitting the NXP SJA1105 driver, the Broadcom b53 and Mediatek mt7530 drivers have been found to apply some VLAN workarounds that are needed in the new driver as well. Therefore this patchset is mostly simply promoting the DSA driver workarounds for VLAN to the generic code. The b53 driver was applying a few workarounds in order to convince DSA that its vlan_filtering setting is not really per-port. This is now simply set by the driver via a DSA variable at probe time. The sja1105 driver will be a second user of this. The mt7530 was also keeping track of when the .port_vlan_filtering callback was being called. Remove the kept state from this driver and simplify dealing with vlan_filtering in the generic case. TODO: Find the best way to deal generically with the situation described below (discussion at https://lkml.org/lkml/2019/4/16/1355): > > +Segregating the switch ports in multiple bridges is supported (e.g. 2 + 2), but > > +all bridges should have the same level of VLAN awareness (either both have > > +``vlan_filtering`` 0, or both 1). Also an inevitable limitation of the fact > > +that VLAN awareness is global at the switch level is that once a bridge with > > +``vlan_filtering`` enslaves at least one switch port, the other un-bridged > > +ports are no longer available for standalone traffic termination. > > That is quite a limitation that I don't think I had fully grasped until > reading your different patches. Since enslaving ports into a bridge > comes after the network device was already made available for use, maybe > you should force the carrier down or something along those lines as soon > as a port is enslaved into a bridge with vlan_filtering=1 to make this > more predictable for the user? ==================== Signed-off-by:
-
