- 07 Nov, 2013 12 commits
-
-
Erik Hugne authored
If appending a received fragment to the pending fragment chain in a unicast link fails, the current code tries to force a retransmission of the fragment by decrementing the 'next received sequence number' field in the link. This is done under the assumption that the failure is caused by an out-of-memory situation, an assumption that does not hold true after the previous patch in this series. A failure to append a fragment can now only be caused by a protocol violation by the sending peer, and it must hence be assumed that it is either malicious or buggy. Either way, the correct behavior is now to reset the link instead of trying to revert its sequence number. So, this is what we do in this commit. Signed-off-by:
Erik Hugne <erik.hugne@ericsson.com> Reviewed-by:
Paul Gortmaker <paul.gortmaker@windriver.com> Signed-off-by:
Jon Maloy <jon.maloy@ericsson.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Erik Hugne authored
When the first fragment of a long data data message is received on a link, a reassembly buffer large enough to hold the data from this and all subsequent fragments of the message is allocated. The payload of each new fragment is copied into this buffer upon arrival. When the last fragment is received, the reassembled message is delivered upwards to the port/socket layer. Not only is this an inefficient approach, but it may also cause bursts of reassembly failures in low memory situations. since we may fail to allocate the necessary large buffer in the first place. Furthermore, after 100 subsequent such failures the link will be reset, something that in reality aggravates the situation. To remedy this problem, this patch introduces a different approach. Instead of allocating a big reassembly buffer, we now append the arriving fragments to a reassembly chain on the link, and deliver the whole chain up to the socket layer once the last fragment has been received. This is safe because the retransmission layer of a TIPC link always delivers packets in strict uninterrupted order, to the reassembly layer as to all other upper layers. Hence there can never be more than one fragment chain pending reassembly at any given time in a link, and we can trust (but still verify) that the fragments will be chained up in the correct order. Signed-off-by:
-
Erik Hugne authored
When a message fragment is received in a broadcast or unicast link, the reception code will append the fragment payload to a big reassembly buffer through a call to the function tipc_recv_fragm(). However, after the return of that call, the logics goes on and passes the fragment buffer to the function tipc_net_route_msg(), which will simply drop it. This behavior is a remnant from the now obsolete multi-cluster functionality, and has no relevance in the current code base. Although currently harmless, this unnecessary call would be fatal after applying the next patch in this series, which introduces a completely new reassembly algorithm. So we change the code to eliminate the redundant call. Signed-off-by:
-
Jonas Jensen authored
The MOXA UC-711X hardware(s) has an ethernet controller that seem to be developed internally. The IC used is "RTL8201CP". This patch adds an MDIO driver which handles the MII bus. Signed-off-by:
Jonas Jensen <jonas.jensen@gmail.com> Signed-off-by:
-
Nikolay Aleksandrov authored
Add new documentation for the packets_per_slave option available for balance-rr mode. Signed-off-by:
Nikolay Aleksandrov <nikolay@redhat.com> Signed-off-by:
-
Nikolay Aleksandrov authored
This patch aims to extend round-robin mode with a new option called packets_per_slave which can have the following values and effects: 0 - choose a random slave 1 (default) - standard round-robin, 1 packet per slave >1 - round-robin when >1 packets have been transmitted per slave The allowed values are between 0 and 65535. This patch also fixes the comment style in bond_xmit_roundrobin(). Signed-off-by:
Veaceslav Falico <vfalico@redhat.com> Signed-off-by:
-
Ursula Braun authored
Check user-defined length in snmp ioctl request and allow request only if it fits into a qeth command buffer. Signed-off-by:
Ursula Braun <ursula.braun@de.ibm.com> Signed-off-by:
Frank Blaschka <frank.blaschka@de.ibm.com> Reviewed-by:
Heiko Carstens <heicars2@linux.vnet.ibm.com> Reported-by:
Nico Golde <nico@ngolde.de> Reported-by:
Fabian Yamaguchi <fabs@goesec.de> Cc: <stable@vger.kernel.org> Signed-off-by:
-
Duan Jiong authored
This patch fixes coccinelle error regarding usage of IS_ERR and PTR_ERR instead of PTR_ERR_OR_ZERO. Signed-off-by:
Duan Jiong <duanj.fnst@cn.fujitsu.com> Signed-off-by:
-
Duan Jiong authored
This patch fixes coccinelle error regarding usage of IS_ERR and PTR_ERR instead of PTR_ERR_OR_ZERO. Signed-off-by:
-
Joe Perches authored
Just an unnecessary semicolon that should be removed... Whitespace neatening of macro too. Signed-off-by:
Joe Perches <joe@perches.com> Signed-off-by:
-
Joe Perches authored
Use a more standard logging style. Convert smsc_<level> macros to use netif_<level>. Remove unused #define PFX Add pr_fmt and neaten pr_<level> uses. Signed-off-by:
-
Joe Perches authored
It's unused, remove it. Signed-off-by:
-
- 06 Nov, 2013 5 commits
-
-
Jason Wang authored
We used to use a percpu structure vq_index to record the cpu to queue mapping, this is suboptimal since it duplicates the work of XPS and loses all other XPS functionality such as allowing user to configure their own transmission steering strategy. So this patch switches to use XPS and suggest a default mapping when the number of cpus is equal to the number of queues. With XPS support, there's no need for keeping per-cpu vq_index and .ndo_select_queue(), so they were removed also. Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Michael S. Tsirkin <mst@redhat.com> Acked-by:
Rusty Russell <rusty@rustcorp.com.au> Acked-by:
Michael S. Tsirkin <mst@redhat.com> Signed-off-by:
Jason Wang <jasowang@redhat.com> Signed-off-by:
-
Duan Jiong authored
Now rt6_alloc_cow() is only called by ip6_pol_route() when rt->rt6i_flags doesn't contain both RTF_NONEXTHOP and RTF_GATEWAY, and rt->rt6i_flags hasn't been changed in ip6_rt_copy(). So there is no neccessary to judge whether rt->rt6i_flags contains RTF_GATEWAY or not. Signed-off-by:
Hannes Frederic Sowa <hannes@stressinduktion.org> Signed-off-by:
-
Hannes Frederic Sowa authored
Commit 1e2bd517 ("udp6: Fix udp fragmentation for tunnel traffic.") changed the calculation if there is enough space to include a fragment header in the skb from a skb->mac_header dervived one to skb_headroom. Because we already peeled off the skb to transport_header this is wrong. Change this back to check if we have enough room before the mac_header. This fixes a panic Saran Neti reported. He used the tbf scheduler which skb_gso_segments the skb. The offsets get negative and we panic in memcpy because the skb was erroneously not expanded at the head. Reported-by:
Saran Neti <Saran.Neti@telus.com> Cc: Pravin B Shelar <pshelar@nicira.com> Signed-off-by:
-
Jason Gunthorpe authored
Commit cc9d4598 'net: mv643xx_eth: use of_phy_connect if phy_node present' made the call to phy_scan optional, if the DT has a link to the phy node. However phy_scan has the side effect of calling phy_addr_set, which writes the phy MDIO address to the ethernet controller. If phy_addr_set is not called, and the bootloader has not set the correct address then the driver will fail to function. Tested on Kirkwood. Signed-off-by:
Jason Gunthorpe <jgunthorpe@obsidianresearch.com> Acked-by:
Sebastian Hesselbarth <sebastian.hesselbarth@gmail.com> Tested-by:
Arnaud Ebalard <arno@natisbad.org> Signed-off-by:
-
Hannes Frederic Sowa authored
Sockets marked with IP_PMTUDISC_INTERFACE won't do path mtu discovery, their sockets won't accept and install new path mtu information and they will always use the interface mtu for outgoing packets. It is guaranteed that the packet is not fragmented locally. But we won't set the DF-Flag on the outgoing frames. Florian Weimer had the idea to use this flag to ensure DNS servers are never generating outgoing fragments. They may well be fragmented on the path, but the server never stores or usees path mtu values, which could well be forged in an attack. (The root of the problem with path MTU discovery is that there is no reliable way to authenticate ICMP Fragmentation Needed But DF Set messages because they are sent from intermediate routers with their source addresses, and the IMCP payload will not always contain sufficient information to identify a flow.) Recent research in the DNS community showed that it is possible to implement an attack where DNS cache poisoning is feasible by spoofing fragments. This work was done by Amir Herzberg and Haya Shulman: <https://sites.google.com/site/hayashulman/files/fragmentation-poisoning.pdf> This issue was previously discussed among the DNS community, e.g. <http://www.ietf.org/mail-archive/web/dnsext/current/msg01204.html>, without leading to fixes. This patch depends on the patch "ipv4: fix DO and PROBE pmtu mode regarding local fragmentation with UFO/CORK" for the enforcement of the non-fragmentable checks. If other users than ip_append_page/data should use this semantic too, we have to add a new flag to IPCB(skb)->flags to suppress local fragmentation and check for this in ip_finish_output. Many thanks to Florian Weimer for the idea and feedback while implementing this patch. Cc: David S. Miller <davem@davemloft.net> Suggested-by:
Florian Weimer <fweimer@redhat.com> Signed-off-by:
-
- 05 Nov, 2013 15 commits
-
-
David S. Miller authored
Bjørn Mork says: ==================== The huawei_cdc_ncm driver. Enrico has been kind enough to let me repost his driver with the changes requested by Oliver Neukum during the last review of this series. The changes I have made from Enricos original v5 series to this version are: v6: - fix to avoid corrupting drvstate->pmcount - fix error return value from huawei_cdc_ncm_suspend() - drop redundant testing for subdriver->suspend during resume - broke a few lines to keep within the 80 columns recommendation - rebased on top of current net-next Enrico's orginal introduction to the v5 series follows below. It explains the background much better than I can. Bjørn [quote Enrico Mioso] So this is a new, revised, edition of the huawei_cdc_ncm.c driver, which supports devices resembling the NCM standard, but using it also as a mean to encapsulate other protocols, as is the case for the Huawei E3131 and E3251 modem devices. Some precisations are needed however - and I encourage discussion on this: and that's why I'm sending this message with a broader CC. Merging those patches might change: - the way Modem Manager interacts with those devices - some regressions might be possible if there are some unknown firmware variants around (Franko?) First of all: I observed the behaviours of two devices. Huawei E3131: this device doesn't accept NDIS setup requests unless they're sent via the embedded AT channel exposed by this driver. So actually we gain funcionality in this case! The second case, is the Huawei E3251: which works with standard NCM driver, still exposing an AT embedded channel. Whith this patch set applied, you gain some funcionality, loosing the ability to catch standard NCM events for now. The device will work in both ways with no problems, but this has to be acknowledged and discussed. Might be we can develop this driver further to change this, when more devices are tested. We where thinking Huawei changed their interfaces on new devices - but probably this driver only works around a nice firmware bug present in E3131, which prevented the modem from being used in NDIS mode. I think committing this is definitely wortth-while, since it will allow for more Huawei devices to be used without serial connection. Some devices like the E3251 also, reports some status information only via the embedded AT channel, at least in my case. Note: I'm not subscribed to any list except the Modem Manager's one, so please CC me, thanks!! [/quote] Enrico Mioso (3): net: cdc_ncm: Export cdc_ncm_{tx,rx}_fixup functions for re-use net: huawei_cdc_ncm: Introduce the huawei_cdc_ncm driver net: cdc_ncm: remove non-standard NCM device IDs ==================== Signed-off-by: -
Enrico Mioso authored
Remove device IDs of NCM-like (but not NCM-conformant) devices, that are handled by the huawwei_cdc_ncm driver now. Signed-off-by:
Enrico Mioso <mrkiko.rs@gmail.com> Signed-off-by:
Bjørn Mork <bjorn@mork.no> Signed-off-by:
-
Enrico Mioso authored
This driver supports devices using the NCM protocol as an encapsulation layer for other protocols, like the E3131 Huawei 3G modem. This drivers approach was heavily inspired by the qmi_wwan/cdc_mbim approach & code model. Signed-off-by:
-
Enrico Mioso authored
Some drivers implementing NCM-like protocols, may re-use those functions, as is the case in the huawei_cdc_ncm driver. Export them via EXPORT_SYMBOL_GPL, in accordance with how other functions have been exported. Signed-off-by:
-
Florent Fourcot authored
The code of flow label in Linux Kernel follows the rules of RFC 1809 (an informational one) for conditions on flow label sharing. There rules are not in the last proposed standard for flow label (RFC 6437), or in the previous one (RFC 3697). Since this code does not follow any current or old standard, we can remove it. With this removal, the ipv6_opt_cmp function is now a dead code and it can be removed too. Changelog to v1: * add justification for the change * remove the condition on IPv6 options [ Remove ipv6_hdr_cmp and it is now unused as well. -DaveM ] Signed-off-by:
Florent Fourcot <florent.fourcot@enst-bretagne.fr> Acked-by:
-
git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless-nextDavid S. Miller authored
John W. Linville says: ==================== Please accept the following pull request intended for the 3.13 tree... I had intended to pass most of these to you as much as two weeks ago. Unfortunately, I failed to account for the effects of bad Internet connections and my own fatique/laziness while traveling. On the bright side, at least these have been baking in linux-next for some time! For the mac80211 bits, Johannes says: "This time I have two fixes for P2P (which requires not using CCK rates) and a workaround for APs with broken WMM information." For the iwlwifi bits, Johannes says: "I have a few fixes for warnings/issues: one from Alex, fixing scan timings, one from Emmanuel fixing a WARN_ON in the DVM driver, one from Stanislaw removing a trigger-happy WARN_ON in the MVM driver and a change from myself to try to recover when the device isn't processing commands quickly." And: "For this round, I have a lot of changes: * power management improvements * BT coexistence improvements/updates * new device support * VHT support * IBSS support (though due to a small bug it requires new firmware) * various other fixes/improvements." For the Bluetooth bits, Gustavo says: "More patches for 3.12, busy times for Bluetooth. More than a 100 commits since the last pull. The bulk of work comes from Johan and Marcel, they are doing fixes and improvements all over the Bluetooth subsystem, as the diffstat can show." For the ath10k and ath6kl bits, Kalle says: "Bartosz added support to ath10k for our 10.x AP firmware branch, which gives us AP specific features and fixes. We still support the main firmware branch as well just like before, ath10k detects runtime what firmware is used. Unfortunately the firmware interface in 10.x branch is somewhat different so there was quite a lot of changes in ath10k for this. Michal and Sujith did some performance improvements in ath10k. Vladimir fixed a compiler warning and Fengguang removed an extra semicolon." For the NFC bits, Samuel says: "It's a fairly big one, with the following highlights: - NFC digital layer implementation: Most NFC chipsets implement the NFC digital layer in firmware, but others have more basic functionalities and expect the host to implement the digital layer. This layer sits below the NFC core. - Sony's port100 support: This is "soft" NFC USB dongle that expects the digital layer to be implemented on the host. This is the first user of our NFC digital stack implementation. - Secure element API: We now provide a netlink API for enabling, disabling and discovering NFC attached (embedded or UICC ones) secure elements. With some userspace help, this allows us to support NFC payments. Only the pn544 driver currently supports that API. - NCI SPI fixes and improvements: In order to support NCI devices over SPI, we fixed and improved our NCI/SPI implementation. The currently most deployed NFC NCI chipset, Broadcom's bcm2079x, supports that mode and we're planning to use our NCI/SPI framework to implement a driver for it. - pn533 fragmentation support in target mode: This was the only missing feature from our pn533 impementation. We now support fragmentation in both Tx and Rx modes, in target mode." On top of all that, brcmfmac and rt2x00 both get the usual flurry of updates. A few other drivers get hit here or there as well. ==================== Signed-off-by: -
Jason Wang authored
Commit 2613af0e (virtio_net: migrate mergeable rx buffers to page frag allocators) try to increase the payload/truesize for MTU-sized traffic. But this will introduce the extra overhead for GSO packets received because of the frag list. This commit tries to reduce this issue by coalesce the possible rx frags when possible during rx. Test result shows the about 15% improvement on full size GSO packet receiving (and even better than before commit 2613af0e). Before this commit: ./netperf -H 192.168.100.4 MIGRATED TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 192.168.100.4 () port 0 AF_INET : demo Recv Send Send Socket Socket Message Elapsed Size Size Size Time Throughput bytes bytes bytes secs. 10^6bits/sec 87380 16384 16384 10.00 20303.87 After this commit: ./netperf -H 192.168.100.4 MIGRATED TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 192.168.100.4 () port 0 AF_INET : demo Recv Send Send Socket Socket Message Elapsed Size Size Size Time Throughput bytes bytes bytes secs. 10^6bits/sec 87380 16384 16384 10.00 23841.26 Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Michael S. Tsirkin <mst@redhat.com> Cc: Michael Dalton <mwdalton@google.com> Cc: Eric Dumazet <edumazet@google.com> Acked-by:
Eric Dumazet <edumazet@google.com> Signed-off-by:
-
Jason Wang authored
Sometimes we need to coalesce the rx frags to avoid frag list. One example is virtio-net driver which tries to use small frags for both MTU sized packet and GSO packet. So this patch introduce skb_coalesce_rx_frag() to do this. Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Michael S. Tsirkin <mst@redhat.com> Cc: Michael Dalton <mwdalton@google.com> Cc: Eric Dumazet <edumazet@google.com> Acked-by:
-
Duan Jiong authored
trivial patch converting ERR_PTR(PTR_ERR()) into ERR_CAST(). No functional changes. Signed-off-by:
-
Jesper Dangaard Brouer authored
As described in commit 5a581b36 (jiffies: Avoid undefined behavior from signed overflow), according to the C standard 3.4.3p3, overflow of a signed integer results in undefined behavior. To fix this, do as the above commit, and do an unsigned subtraction, and interpreting the result as a signed two's-complement number. This is based on the theory from RFC 1982 and is nicely described in wikipedia here: https://en.wikipedia.org/wiki/Serial_number_arithmetic#General_Solution A side-note, I have seen practical issues with the previous logic when dealing with 16-bit, on a 64-bit machine (gcc version 4.4.5). This were 32-bit, which I have not observed issues with. Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Signed-off-by:
Jesper Dangaard Brouer <netoptimizer@brouer.com> Acked-by:
Paul E. McKenney <paulmck@linux.vnet.ibm.com> Signed-off-by:
-
git://gitorious.org/linux-can/linux-can-nextDavid S. Miller authored
Marc Kleine-Budde says: ==================== here's a pull request for net-next. It includes a patch by Oliver Hartkopp et al. that adds documentation for the broadcast manager to Documentation/networking/can.txt. Three patches by me that clean up the netlink handling code in the CAN core. And another patch that removes a not needed function from the ti_hecc driver. ==================== Signed-off-by: -
Yuchung Cheng authored
Slow start now increases cwnd by 1 if an ACK acknowledges some packets, regardless the number of packets. Consequently slow start performance is highly dependent on the degree of the stretch ACKs caused by receiver or network ACK compression mechanisms (e.g., delayed-ACK, GRO, etc). But slow start algorithm is to send twice the amount of packets of packets left so it should process a stretch ACK of degree N as if N ACKs of degree 1, then exits when cwnd exceeds ssthresh. A follow up patch will use the remainder of the N (if greater than 1) to adjust cwnd in the congestion avoidance phase. In addition this patch retires the experimental limited slow start (LSS) feature. LSS has multiple drawbacks but questionable benefit. The fractional cwnd increase in LSS requires a loop in slow start even though it's rarely used. Configuring such an increase step via a global sysctl on different BDPS seems hard. Finally and most importantly the slow start overshoot concern is now better covered by the Hybrid slow start (hystart) enabled by default. Signed-off-by:
Yuchung Cheng <ycheng@google.com> Signed-off-by:
Neal Cardwell <ncardwell@google.com> Signed-off-by:
-
Yuchung Cheng authored
Applications have started to use Fast Open (e.g., Chrome browser has such an optional flag) and the feature has gone through several generations of kernels since 3.7 with many real network tests. It's time to enable this flag by default for applications to test more conveniently and extensively. Signed-off-by:
-
git://git.kernel.org/pub/scm/linux/kernel/git/pablo/nftablesDavid S. Miller authored
Pablo Neira Ayuso says: ==================== This batch contains fives nf_tables patches for your net-next tree, they are: * Fix possible use after free in the module removal path of the x_tables compatibility layer, from Dan Carpenter. * Add filter chain type for the bridge family, from myself. * Fix Kconfig dependencies of the nf_tables bridge family with the core, from myself. * Fix sparse warnings in nft_nat, from Tomasz Bursztyka. * Remove duplicated include in the IPv4 family support for nf_tables, from Wei Yongjun. ==================== Signed-off-by: -
git://git.kernel.org/pub/scm/linux/kernel/git/pablo/nf-nextDavid S. Miller authored
Pablo Neira Ayuso says: ==================== This is another batch containing Netfilter/IPVS updates for your net-next tree, they are: * Six patches to make the ipt_CLUSTERIP target support netnamespace, from Gao feng. * Two cleanups for the nf_conntrack_acct infrastructure, introducing a new structure to encapsulate conntrack counters, from Holger Eitzenberger. * Fix missing verdict in SCTP support for IPVS, from Daniel Borkmann. * Skip checksum recalculation in SCTP support for IPVS, also from Daniel Borkmann. * Fix behavioural change in xt_socket after IP early demux, from Florian Westphal. * Fix bogus large memory allocation in the bitmap port set type in ipset, from Jozsef Kadlecsik. * Fix possible compilation issues in the hash netnet set type in ipset, also from Jozsef Kadlecsik. * Define constants to identify netlink callback data in ipset dumps, again from Jozsef Kadlecsik. * Use sock_gen_put() in xt_socket to replace xt_socket_put_sk, from Eric Dumazet. * Improvements for the SH scheduler in IPVS, from Alexander Frolkin. * Remove extra delay due to unneeded rcu barrier in IPVS net namespace cleanup path, from Julian Anastasov. * Save some cycles in ip6t_REJECT by skipping checksum validation in packets leaving from our stack, from Stanislav Fomichev. * Fix IPVS_CMD_ATTR_MAX definition in IPVS, larger that required, from Julian Anastasov. ==================== Signed-off-by:
-
- 04 Nov, 2013 8 commits
-
-
Dan Carpenter authored
We need to use the _safe version of list_for_each_entry() here otherwise we have a use after free bug. Signed-off-by:
Dan Carpenter <dan.carpenter@oracle.com> Signed-off-by:
Pablo Neira Ayuso <pablo@netfilter.org>
-
git://git.kernel.org/pub/scm/linux/kernel/git/jesse/openvswitchDavid S. Miller authored
Jesse Gross says: ==================== Open vSwitch A set of updates for net-next/3.13. Major changes are: * Restructure flow handling code to be more logically organized and easier to read. * Rehashing of the flow table is moved from a workqueue to flow installation time. Before, heavy load could block the workqueue for excessive periods of time. * Additional debugging information is provided to help diagnose megaflows. * It's now possible to match on TCP flags. ==================== Signed-off-by: -
David S. Miller authored
Or Gerlitz says: ==================== Mellanox driver updates This patch set from Jack Morgenstein does the following: 1. Fix MAC/VLAN SRIOV implementation, and add wrapper functions for VLAN allocation and de-allocation (patches 1-6). 2. Implements resource quotas when running under SRIOV (patches 7-10). Patch 7 is a small bug fix, and patches 8-10 implement the quotas. Quotas are implemented per resource type for VFs and the PF, to prevent any entity from simply grabbing all the resources for itself and leaving the other entities unable to obtain such resources. The series is against net-next commit ba486502 "ipv6: remove the unnecessary statement in find_match()" changes from V0: - dropped the 1st patch which needs to go to -stable and hence through net, not net-next ==================== Signed-off-by:
-
Jack Morgenstein authored
Implements resource quota grant decision when resources are requested, for the following resources: QPs, CQs, SRQs, MPTs, MTTs, vlans, MACs, and Counters. When granting a resource, the quota system increases the allocated-count for that slave. When the slave later frees the resource, its allocated-count is reduced. A spinlock is used to protect the integrity of each resource's free-pool counter. (One slave may be in the process of being granted a resource while another slave has crashed, initiating cleanup of that slave's resource quotas). Signed-off-by:
Jack Morgenstein <jackm@dev.mellanox.co.il> Signed-off-by:
Or Gerlitz <ogerlitz@mellanox.com> Signed-off-by:
-
Jack Morgenstein authored
In current kernels, the mlx4 driver running on a VM does not differentiate between max resource numbers for the HCA and max quotas -- it simply takes the quota values passed to it as max-resource values. However, the driver actually requires the VFs to be aware of the actual number of resources that the HCA was initialized with, for QPs, CQs, SRQs and MPTs. For QPs, CQs and SRQs, the reason is that in completion handling the driver must know which of the 24 bits are the actual resource number, and which are "padding" bits. For MPTs, also, the driver assumes knowledge of the number of MPTs in the system. The previous commit fixes the quota logic on the VM for the quota values passed to it by QUERY_FUNC_CAPS. For QPs, CQs, SRQs, and MPTs, it takes the max resource numbers from QUERY_HCA (and not QUERY_FUNC_CAPS). The quotas passed in QUERY_FUNC_CAPS are used to report max resource number values in the response to ib_query_device. However, the Hypervisor driver must consider that VMs may be running previous kernels, and compatibility must be preserved. To resolve the incompatibility with previous kernels running on VMs, we deprecated the quota fields in mlx4_QUERY_FUNC_CAP. In the deprecated fields, we pass the max-resource values from INIT_HCA The quota fields are moved to a new location, and the current kernel driver takes the proper values from that location. There is also a new flag in dword 0, bit 28 of the mlx4_QUERY_FUNC_CAP mailbox; if this flag is set, the (VM) driver takes the quota values from the new location. VMs running previous kernels will work properly, except that the max resource numbers reported in ib_query_device for these resources will be too high. The Hypervisor driver will, however, enforce the quotas for these VMs. Signed-off-by:
-
Jack Morgenstein authored
This is step #1 for implementing SRIOV resource quotas for VFs. Quotas are implemented per resource type for VFs and the PF, to prevent any entity from simply grabbing all the resources for itself and leaving the other entities unable to obtain such resources. Resources which are allocated using quotas: QPs, CQs, SRQs, MPTs, MTTs, MAC, VLAN, and Counters. The quota system works as follows: Each entity (VF or PF) is given a max number of a given resource (its quota), and a guaranteed minimum number for each resource (starvation prevention). For QPs, CQs, SRQs, MPTs and MTTs: 50% of the available quantity for the resource is divided equally among the PF and all the active VFs (i.e., the number of VFs in the mlx4_core module parameter "num_vfs"). This 50% represents the "guaranteed minimum" pool. The other 50% is the "free pool", allocated on a first-come-first-serve basis. For each VF/PF, resources are first allocated from its "guaranteed-minimum" pool. When that pool is exhausted, the driver attempts to allocate from the resource "free-pool". The quota (i.e., max) for the VFs and the PF is: The free-pool amount (50% of the real max) + the guaranteed minimum For MACs: Guarantee 2 MACs per VF/PF per port. As a result, since we have only 128 MACs per port, reduce the allowable number of VFs from 64 to 63. Any remaining MACs are put into a free pool. For VLANs: For the PF, the per-port quota is 128 and guarantee is 64 (to allow the PF to register at least a VLAN per VF in VST mode). For the VFs, the per-port quota is 64 and the guarantee is 0. We assume that VGT VFs are trusted not to abuse the VLAN resource. For Counters: For all functions (PF and VFs), the quota is 128 and the guarantee is 0. In this patch, we define the needed structures, which are added to the resource-tracker struct. In addition, we do initialization for the resource quota, and adjust the query_device response to use quotas rather than resource maxima. As part of the implementation, we introduce a new field in mlx4_dev: quotas. This field holds the resource quotas used to report maxima to the upper layers (ib_core, via query_device). The HCA maxima of these values are passed to the VFs (via QUERY_HCA) so that they may continue to use these in handling QPs, CQs, SRQs and MPTs. Signed-off-by: -
Jack Morgenstein authored
In procedure mlx4_init_mr_table(), slaves should do no processing, but should return success. This initialization is hypervisor-only. However, the check for num_mpts being a power-of-2 was performed before the check to return immediately if the driver is for a slave. This resulted in spurious failures. The order of performing the checks is reversed, so that if the driver is for a slave, no processing is done and success is returned. Signed-off-by:
-
Jack Morgenstein authored
In upstream kernels under SRIOV, the vlan register/unregister calls were NOPs (doing nothing and returning OK). We detect these old calls from guests (via the comm channel), since previously the port number in mlx4_register_vlan was passed (improperly) in the out_param. This has been corrected so that the port number is now passed in bits 8..15 of the in_modifier field. For old calls, these bits will be zero, so if the passed port number is zero, we can still look at the out_param field to see if it contains a valid port number. If yes, the VM is running an old driver. Since for old drivers, the register/unregister_vlan wrappers were NOPs, we continue this policy -- the reason being that upstream had an additional bug in eth driver running on guests (where procedure mlx4_en_vlan_rx_kill_vid() had the following code: if (!mlx4_find_cached_vlan(mdev->dev, priv->port, vid, &idx)) mlx4_unregister_vlan(mdev->dev, priv->port, idx); else en_err(priv, "could not find vid %d in cache\n", vid); On a VM, mlx4_find_cached_vlan() will always fail, since the vlan cache is located on the Hypervisor; on guests it is empty. Therefore, if we allow upstream guests to register vlans, we will have vlan leakage since the unregister will never be performed. Leaving vlan reg/unreg for old guest drivers as a NOP is not a feature regression, since in upstream the register/unregister vlan wrapper is a NOP. Signed-off-by:
-
