- 05 Jan, 2018 20 commits
-
-
David S. Miller authored
Sowmini Varadhan says: ==================== rds: use RCU between work-enqueue and connection teardown This patchset follows up on the root-cause mentioned in https://www.spinics.net/lists/netdev/msg472849.html Patch1 implements some code refactoring that was suggeseted as an enhancement in http://patchwork.ozlabs.org/patch/843157/ It replaces the c_destroy_in_prog bit in rds_connection with an atomically managed flag in rds_conn_path. Patch2 builds on Patch1 and uses RCU to make sure that work is only enqueued if the connection destroy is not already in progress: the test-flag-and-enqueue is done under rcu_read_lock, while destroy first sets the flag, uses synchronize_rcu to wait for existing reader threads to complete, and then starts all the work-cancellation. Since I have not been able to reproduce the original stack traces reported by syszbot, and these are fixes for a race condition that are based on code-inspection I am not marking these as reported-by at this time. ==================== Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Sowmini Varadhan authored
rds_sendmsg() can enqueue work on cp_send_w from process context, but it should not enqueue this work if connection teardown has commenced (else we risk enquing work after rds_conn_path_destroy() has assumed that all work has been cancelled/flushed). Similarly some other functions like rds_cong_queue_updates and rds_tcp_data_ready are called in softirq context, and may end up enqueuing work on rds_wq after rds_conn_path_destroy() has assumed that all workqs are quiesced. Check the RDS_DESTROY_PENDING bit and use rcu synchronization to avoid all these races. Signed-off-by:
Sowmini Varadhan <sowmini.varadhan@oracle.com> Acked-by:
Santosh Shilimkar <santosh.shilimkar@oracle.com> Signed-off-by:
-
Sowmini Varadhan authored
Replace c_destroy_in_prog by using a bit in cp_flags that can set/tested atomically. Signed-off-by:
-
David S. Miller authored
Jon Maloy says: ==================== tipc: two small cleanups These two commits are based on commit f9c935db ("tipc: fix problems with multipoint-to-point flow control") which has been applied to 'net' but not yet to 'net-next'. ==================== Signed-off-by:
-
Jon Maloy authored
We simplify the sorting algorithm in tipc_update_member(). We also make the remaining conditional call to this function unconditional, since the same condition now is tested for inside the said function. Acked-by:
Ying Xue <ying.xue@windriver.com> Signed-off-by:
Jon Maloy <jon.maloy@ericsson.com> Signed-off-by:
-
Jon Maloy authored
We rename some functions and variables, to make their purpose clearer. - tipc_group::congested -> tipc_group::small_win. Members in this list are not necessarily (and typically) congested. Instead, they may *potentially* be subject to congestion because their send window is less than ADV_IDLE, and therefore need to be checked during message transmission. - tipc_group_is_receiver() -> tipc_group_is_sender(). This socket will accept messages coming from members fulfilling this condition, i.e., they are senders from this member's viewpoint. - tipc_group_is_enabled() -> tipc_group_is_receiver(). Members fulfilling this condition will accept messages sent from the current socket, i.e., they are receivers from its viewpoint. There are no functional changes in this commit. Acked-by:
-
Wei Yongjun authored
Fix to return error code -ENODEV from the chip not found error handling case instead of 0(ret have been overwritten to 0 by lan9303_read()), as done elsewhere in this function. Signed-off-by:
Wei Yongjun <weiyongjun1@huawei.com> Reviewed-by:
Egil Hjelmeland <privat@egil-hjelmeland.no> Signed-off-by:
-
David S. Miller authored
Florian Fainelli says: ==================== net: dsa: Move padding into Broadcom tagger This patch series moves the padding of short packets to where it belongs within the DSA Broadcom tagger code, I just found myself doing this for a third driver, which was a clear indication this was wrong and did not scale. ==================== Signed-off-by: -
Florian Fainelli authored
DSA now correctly pads short packets within net/dsa/tag_brcm.c such that this it is no longer necessary to do this within bgmac. Signed-off-by:
Florian Fainelli <f.fainelli@gmail.com> Signed-off-by:
-
Florian Fainelli authored
Short packet padding added to the driver is only necessary when using Broadcom tags, but since this is now taken care of net/dsa/tag_brcm.c, we are guaranteed being given correctly padded packets. Signed-off-by:
-
Florian Fainelli authored
Instead of having the different master network device drivers potentially used by DSA/Broadcom tags, move the padding necessary for the switches to accept short packets where it makes most sense: within tag_brcm.c. This avoids multiplying the number of similar commits to e.g: bgmac, bcmsysport, etc. Signed-off-by:
-
Quentin Monnet authored
The definition of functions tcf_block_get() and tcf_block_get_ext() depends of CONFIG_NET_CLS being set. When those functions gained extack support, only one version of the declaration of those functions was updated. Function tcf_block_get() was later fixed with commit 3c149091 ("net: sch: api: fix tcf_block_get"). Change arguments of tcf_block_get_ext() for the case when CONFIG_NET_CLS is not set. Fixes: 8d1a77f9 ("net: sch: api: add extack support in tcf_block_get") Signed-off-by:
Quentin Monnet <quentin.monnet@netronome.com> Acked-by:
Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by:
-
Soheil Hassas Yeganeh authored
On multi-threaded processes, one common architecture is to have one (or a small number of) threads polling sockets, and a considerably larger pool of threads reading form and writing to the sockets. When we set RPS core on tcp_poll() or udp_poll() we essentially steer all packets of all the polled FDs to one (or small number of) cores, creaing a bottleneck and/or RPS misprediction. Another common architecture is to shard FDs among threads pinned to cores. In such a setting, setting RPS core in tcp_poll() and udp_poll() is redundant because the RFS core is correctly set in recvmsg and sendmsg. Thus, revert the following commit: c3f1dbaf ("net: Update RFS target at poll for tcp/udp"). Signed-off-by:
Soheil Hassas Yeganeh <soheil@google.com> Signed-off-by:
Willem de Bruijn <willemb@google.com> Signed-off-by:
Eric Dumazet <edumazet@google.com> Signed-off-by:
Neal Cardwell <ncardwell@google.com> Signed-off-by:
-
Soheil Hassas Yeganeh authored
We should only record RPS on normal reads and writes. In single threaded processes, all calls record the same state. In multi-threaded processes where a separate thread processes errors, the RFS table mispredicts. Note that, when CONFIG_RPS is disabled, sock_rps_record_flow is a noop and no branch is added as a result of this patch. Signed-off-by:
-
David S. Miller authored
James Chapman says: ==================== l2tp: remove configurable offset parameters This patch series removes all code to support a configurable offset in transmitted l2tp packets. Code to handle this is incomplete and buggy and has been this way for years. If anyone tried to configure an offset, it would be ignored for L2TPv2 tunnels, or for L2TPv3 tunnels, could result in L2TPv3 packets being transmitted which are not compliant with L2TPv3 RFC3931. This patch series removes the support for configurable offsets. No known userspace l2tp daemon configures an offset. However, iproute2's "ip l2tp" command has an offset parameter and if set, the value is passed to the kernel. This is the most likely use case where offsets might be configured, e.g. ip l2tp add tunnel local 1.1.1.1 remote 1.1.1.2 tunnel_id 1 \ peer_tunnel_id 2 encap ip ip l2tp add session name l2tp0 tunnel_id 1 session_id 1 \ peer_session_id 2 offset 8 The above would result in packets being transmitted to 1.1.1.2 with 8 bytes padding between the L2TPv3 header and the payload. The peer would need to be configured with the same offset value. However, the packets are not compliant with the L2TPv3 RFC, hence I think it's unlikely that offset is being used. With this patch series applied, the offset would not be configured. The peer would need to be modified to remove its offset setting too. iproute2 should be modified to remove or ignore the ip l2tp offset parameter. This issue was discovered when reviewing a patch series from lorenzo.bianconi@redhat.com which adds another netlink attribute to configure the expected offset in received L2TPv3 packets. This change is reverted by this series because offsets do not exist in L2TPv3 packets. These commits are: commit f15bc54e ("l2tp: add peer_offset parameter") commit 820da535 ("l2tp: fix missing print session offset info") In more detail: The L2TPv2 protocol supports a variable offset from the L2TPv2 header to the payload to give the sender implementation some flexibility for data alignment when adding L2TP headers on to payloads. The offset value is indicated by an optional field in the L2TP header. Our L2TP implementation already detects the presence of the optional offset in received packets and skips those bytes when parsing packets. All transmitted L2TPv2 packets are always transmitted with no offset. L2TPv3 has no optional offset field in the L2TPv3 packet header. Instead, L2TPv3 defines optional fields in a "Layer-2 Specific Sublayer". At the time when the original L2TP code was written, there was talk at IETF of offset being implemented in a new Layer-2 Specific Sublayer. A L2TP_ATTR_OFFSET netlink attribute was added so that this offset could be configured and the intention was to allow it to be also used to set the tx offset for L2TPv2. However, no L2TPv3 offset was ever specified and the L2TP_ATTR_OFFSET parameter was forgotten about. Setting L2TP_ATTR_OFFSET results in L2TPv3 packets being transmitted with the specified number of bytes padding between L2TPv3 header and payload. This is not compliant with L2TPv3 RFC3931. So this change removes the configurable offset altogether while retaining L2TP_ATTR_OFFSET in the API for backwards compatibility. If L2TP_ATTR_OFFSET is given, its value is now silently ignored. ==================== Reviewed-by:Guillaume Nault <g.nault@alphalink.fr> Tested-by:
-
James Chapman authored
Signed-off-by:
James Chapman <jchapman@katalix.com> Signed-off-by:
-
James Chapman authored
If L2TP_ATTR_OFFSET is set to a non-zero value in L2TPv3 tunnels, it results in L2TPv3 packets being transmitted which might not be compliant with the L2TPv3 RFC. This patch has l2tp ignore the offset setting and send all packets with no offset. In more detail: L2TPv2 supports a variable offset from the L2TPv2 header to the payload. The offset value is indicated by an optional field in the L2TP header. Our L2TP implementation already detects the presence of the optional offset and skips that many bytes when handling data received packets. All transmitted packets are always transmitted with no offset. L2TPv3 has no optional offset field in the L2TPv3 packet header. Instead, L2TPv3 defines optional fields in a "Layer-2 Specific Sublayer". At the time when the original L2TP code was written, there was talk at IETF of offset being implemented in a new Layer-2 Specific Sublayer. A L2TP_ATTR_OFFSET netlink attribute was added so that this offset could be configured and the intention was to allow it to be also used to set the tx offset for L2TPv2. However, no L2TPv3 offset was ever specified and the L2TP_ATTR_OFFSET parameter was forgotten about. Setting L2TP_ATTR_OFFSET results in L2TPv3 packets being transmitted with the specified number of bytes padding between L2TPv3 header and payload. This is not compliant with L2TPv3 RFC3931. This change removes the configurable offset altogether while retaining L2TP_ATTR_OFFSET for backwards compatibility. Any L2TP_ATTR_OFFSET value is ignored. Signed-off-by:
-
James Chapman authored
Revert commit 820da535 ("l2tp: fix missing print session offset info"). The peer_offset parameter is removed. Signed-off-by:
-
James Chapman authored
Revert commit f15bc54e ("l2tp: add peer_offset parameter"). This is removed because it is adding another configurable offset and configurable offsets are being removed. Signed-off-by:
-
Arnd Bergmann authored
The uplink_rpriv variable was added at the start of the function but only used inside of an #ifdef: drivers/net/ethernet/mellanox/mlx5/core/en_tc.c: In function 'mlx5e_route_lookup_ipv6': drivers/net/ethernet/mellanox/mlx5/core/en_tc.c:1549:25: error: unused variable 'uplink_rpriv' [-Werror=unused-variable] This moves the declaration into that #ifdef as well. Fixes: 5ed99fb4 ("net/mlx5e: Move ethernet representors data into separate struct") Signed-off-by:
Arnd Bergmann <arnd@arndb.de> Acked-by:
Saeed Mahameed <saeedm@mellanox.com> Signed-off-by:
-
- 04 Jan, 2018 6 commits
-
-
David S. Miller authored
Merge tag 'mac80211-next-for-davem-2018-01-04' of git://git.kernel.org/pub/scm/linux/kernel/git/jberg/mac80211-next Johannes Berg says: ==================== We have things all over the place, no point listing them. One thing is notable: I applied two patches and later reverted them - we'll get back to that once all the driver situation is sorted out. ==================== Signed-off-by:
-
Prashant Sreedharan authored
This patch adds the support for Macronix NVRAM Signed-off-by:
Prashant Sreedharan <prashant.sreedharan@broadcom.com> Signed-off-by:
Satish Baddipadige <satish.baddipadige@broadcom.com> Reviewed-by:
Michael Chan <michael.chan@broadcom.com> Signed-off-by:
-
David S. Miller authored
Egil Hjelmeland says: ==================== net: dsa: lan9303: phy_addr_sel_strap rename and retype Non functional cleanups involving chip->phy_addr_sel_strap. As promised in https://lkml.org/lkml/2017/11/6/273 ==================== Signed-off-by:
-
Egil Hjelmeland authored
Simplify calculation of chip->phy_addr_base in lan9303_detect_phy_setup(). Use GENMASK to calculate phys_mii_mask from LAN9303_NUM_PORTS and phy_addr_base. Signed-off-by:
-
Egil Hjelmeland authored
chip->phy_addr_sel_strap is declared as a bool, but is also used as an integer address base. Rename 'phy_addr_sel_strap' to 'phy_addr_base', and change type to int. Signed-off-by:
-
Peter Große authored
Instead of calling ieee80211_recalc_txpower on monitor interfaces directly, call it using the virtual monitor interface, if one exists. In case of a single monitor interface given, reject setting TX power, if no virtual monitor interface exists. That being checked, don't warn in ieee80211_bss_info_change_notify, after setting TX power on a monitor interface. Fixes warning: ------------[ cut here ]------------ WARNING: CPU: 0 PID: 2193 at net/mac80211/driver-ops.h:167 ieee80211_bss_info_change_notify+0x111/0x190 Modules linked in: uvcvideo videobuf2_vmalloc videobuf2_memops videobuf2_v4l2 videobuf2_core rndis_host cdc_ether usbnet mii tp_smapi(O) thinkpad_ec(O) ohci_hcd vboxpci(O) vboxnetadp(O) vboxnetflt(O) v boxdrv(O) x86_pkg_temp_thermal kvm_intel kvm irqbypass iwldvm iwlwifi ehci_pci ehci_hcd tpm_tis tpm_tis_core tpm CPU: 0 PID: 2193 Comm: iw Tainted: G O 4.12.12-gentoo #2 task: ffff880186fd5cc0 task.stack: ffffc90001b54000 RIP: 0010:ieee80211_bss_info_change_notify+0x111/0x190 RSP: 0018:ffffc90001b57a10 EFLAGS: 00010246 RAX: 0000000000000006 RBX: ffff8801052ce840 RCX: 0000000000000064 RDX: 00000000fffffffc RSI: 0000000000040000 RDI: ffff8801052ce840 RBP: ffffc90001b57a38 R08: 0000000000000062 R09: 0000000000000000 R10: ffff8802144b5000 R11: ffff880049dc4614 R12: 0000000000040000 R13: 0000000000000064 R14: ffff8802105f0760 R15: ffffc90001b57b48 FS: 00007f92644b4580(0000) GS:ffff88021e200000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00007f9263c109f0 CR3: 00000001df850000 CR4: 00000000000406f0 Call Trace: ieee80211_recalc_txpower+0x33/0x40 ieee80211_set_tx_power+0x40/0x180 nl80211_set_wiphy+0x32e/0x950 Reported-by:
Peter Große <pegro@friiks.de> Signed-off-by:
Johannes Berg <johannes.berg@intel.com>
-
- 03 Jan, 2018 14 commits
-
-
David S. Miller authored
Jakub Kicinski says: ==================== nfp: flower: repr link state Dirk says: This series provides two updates towards the link state of reprs in the flower nfp app. Patch #1 improves the way link state is reported for reprs. Instead of starting with an assumed 'UP' state, always assume the link state is 'DOWN' and then modify this only on events received from firmware. Patch #2 adds a new nfp_app hook, repr_preclean. This callback is executed before reprs are removed from the app context and is executed per repr. Patch #3 implements the new REIFY control message, used to indicate when reprs are created and destroyed. Firmware uses these messages to prevent communication about any particular port when the driver doesn't know about the repr yet or anymore. ==================== Signed-off-by: -
Dirk van der Merwe authored
The PORT_REIFY message indicates whether reprs have been created or when they are about to be destroyed. This is necessary so firmware can know which state the driver is in, e.g. the firmware will not send any control messages related to ports when the reprs are destroyed. This prevents nuisance warning messages printed whenever the firmware sends updates for non-existent reprs. Signed-off-by:
Dirk van der Merwe <dirk.vandermerwe@netronome.com> Reviewed-by:
Jakub Kicinski <jakub.kicinski@netronome.com> Signed-off-by:
-
Dirk van der Merwe authored
Just before a repr is cleaned up, we give the app a chance to perform some preclean configuration while the reprs pointer is still configured for the app. Signed-off-by:
-
Dirk van der Merwe authored
Instead of starting up reprs assuming that there is link, only respond to the link state reported by firmware. Furthermore, ensure link is down after repr netdevs are created. Signed-off-by:
-
Andrew Lunn authored
Ethernet switch on the MDIO bus have historically performed their own handling of the GPIO reset line. The resent patch to have the MDIO core handle the reset has broken the switch drivers, in that they cannot claim the GPIO. Some switch drivers need more control over the GPIO line than what the MDIO core provides. So restore the historical behaviour by only performing a reset of PHYs, not switches. Fixes: bafbdd52 ("phylib: Add device reset GPIO support") Reported-by:
Sean Wang <sean.wang@mediatek.com> Signed-off-by:
Andrew Lunn <andrew@lunn.ch> Reviewed-by:
Geert Uytterhoeven <geert+renesas@glider.be> Tested-by:
-
David S. Miller authored
Russell King says: ==================== Resolve races in phy accessors This series resolves races with various accesses to PHY registers. The first five patches are necessary before we add phylink support to mvneta, the remaining three are merely cleanups for unobserved races, and hence are less critical. There are two possible classes of races that can occur: where we write to a page register that changes the meaning of a group of other registers, and where we read-modify-write a register. Resolve these races by performing the accesses under the mdio bus lock, ensuring that no other user can access the bus while the series of atomic operations are being performed. These patches have been posted before, and have been modified along the lines of previous feedback: - The third patch was originally reviewed by Florian, but as I've added __phy_modify() to it, I've removed that attributation. - Included generic page-based accessors as suggested last time around. - Since we have the unlocked __phy_modify() in this patch series, it is sensible to include the changes for this to marvell.c - these accessors have to change anyway to avoid deadlocks on the mdio bus lock. I haven't been able to test the at803x.c changes yet beyond compile testing - although I do have systems with an ar8035 PHY. However, they should be straight forward to review. This is targetted for net-next because the races have not been found in existing drivers, but have been observed with phylink integrated into mvneta - that's not to say that the races do not exist today, they are just unobserved (probably through lack of rigorous enough testing.) The race provoking condition is detailed in patch 5. ==================== Signed-off-by: -
Russell King authored
Convert read-modify-write sequences in at803x, Marvell and core phylib to use phy_modify() to ensure safety. Signed-off-by:
Russell King <rmk+kernel@armlinux.org.uk> Reviewed-by:
-
Russell King authored
Add phy_modify() convenience accessor to complement the mdiobus counterpart. Signed-off-by:
-
Russell King authored
For paged accesses to be truely safe, we need to hold the bus lock to prevent anyone else gaining access to the registers while we modify them. The phydev->lock mutex does not do this: userspace via the MII ioctl can still sneak in and read or write any register while we are on a different page, and the suspend/resume methods can be called by a thread different to the thread polling the phy status. Races have been observed with mvneta on SolidRun Clearfog with phylink, particularly between the phylib worker reading the PHYs status, and the thread resuming mvneta, calling phy_start() which then calls through to m88e1121_config_aneg_rgmii_delays(), which tries to read-modify-write the MSCR register: CPU0 CPU1 marvell_read_status_page() marvell_set_page(phydev, MII_MARVELL_FIBER_PAGE) ... m88e1121_config_aneg_rgmii_delays() set_page(MII_MARVELL_MSCR_PAGE) phy_read(phydev, MII_88E1121_PHY_MSCR_REG) marvell_set_page(phydev, MII_MARVELL_COPPER_PAGE); ... phy_write(phydev, MII_88E1121_PHY_MSCR_REG) The result of this is we end up writing the copper page register 21, which causes the copper PHY to be disabled, and the link partner sees the link immediately go down. Solve this by taking the bus lock instead of the PHY lock, thereby preventing other accesses to the PHY while we are accessing other PHY pages. Signed-off-by:
-
Russell King authored
Add a set of paged phy register accessors which are inherently safe in their design against other accesses interfering with the paged access. Signed-off-by:
-
Russell King authored
Add unlocked versions of the bus accessors, which allows access to the bus with all the tracing. These accessors validate that the bus mutex is held, which is a basic requirement for all mii bus accesses. Also added is a read-modify-write unlocked accessor with the same locking requirements. Signed-off-by:
-
Russell King authored
Use unlocked accessors for indirect MMD accesses to clause 22 PHYs. This permits tracing of these accesses. Reviewed-by:
-
Russell King authored
Add unlocked versions of the bus accessors, which allows access to the bus with all the tracing. These accessors validate that the bus mutex is held, which is a basic requirement for all mii bus accesses. Reviewed-by:
-
Rahul Lakkireddy authored
Collect TX rate limiting related information in UP CIM logs. Signed-off-by:
Rahul Lakkireddy <rahul.lakkireddy@chelsio.com> Signed-off-by:
Ganesh Goudar <ganeshgr@chelsio.com> Signed-off-by:
-
