Re: network locks up with udp traffic

> On 23 янв. 2015 г., at 21:58, Adrian Chadd wrote: > > Hi, > > When this next happens, please do this as root: > > procstat -ka > > That'll hopefully provide enough information to figure out which > processes are blocking where and how they got there. > > Hello, I am attaching procstat

[Bug 191520] [PATCH] Implementation of draft-ieft-tcpm-newcwv-06

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=191520 Kubilay Kocak changed: What|Removed |Added Keywords||feature, needs-qa, patch -- You a

Re: network locks up with udp traffic

Hi, Can you compile your kernel with WITNESS and repeat this test? It looks like it's stuck in some lock ordering thing between some UDP paths and an inpcb lookup. It should log in dmesg the first time a LOR occurs, regardless of whether it actively causes a hang. (And yes, this is definitely cr

Re: network locks up with udp traffic

> On 24 янв. 2015 г., at 20:29, Adrian Chadd wrote: > > Hi, > > Can you compile your kernel with WITNESS and repeat this test? It > looks like it's stuck in some lock ordering thing between some UDP > paths and an inpcb lookup. > > It should log in dmesg the first time a LOR occurs, regardless

Re: network locks up with udp traffic

Hi! To be clear: * is your kernel modified in any way; and * did witness give you a full stacktrace as part of the lock order reversal? All of that would be good. Thanks, On 24 January 2015 at 09:53, Dmitry Sivachenko wrote: > >> On 24 янв. 2015 г., at 20:29, Adrian Chadd wrote: >> >> Hi,

Re: network locks up with udp traffic

> On 24 янв. 2015 г., at 21:56, Adrian Chadd wrote: > > Hi! > > To be clear: > > * is your kernel modified in any way; and No, it isn't > * did witness give you a full stacktrace as part of the lock order > reversal? All of that would be good. > I provided the full output I saw on conso

[Bug 188897] [dc] dc ethernet driver seems to prevent the detection of other NIC chipsets

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=188897 --- Comment #7 from c...@eugeneweb.com --- Created attachment 152106 --> https://bugs.freebsd.org/bugzilla/attachment.cgi?id=152106&action=edit This is a verbose dmesg for another dc quadport card, a DFE-570TX -- You are receiving this m

[Bug 194515] Fatal Trap 12 Kernel with vimage

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=194515 --- Comment #19 from Craig Rodrigues --- (In reply to darius from comment #18) I'm OK with merging this to stable/10, but I would like to get some feedback from PF users on the state of things in CURRENT before merging. What is your experi

Re: network locks up with udp traffic

Ah, try adding STACK to your config? -a On 24 January 2015 at 11:45, Dmitry Sivachenko wrote: > >> On 24 янв. 2015 г., at 21:56, Adrian Chadd wrote: >> >> Hi! >> >> To be clear: >> >> * is your kernel modified in any way; and > > > No, it isn't > > >> * did witness give you a full stacktrace

[Differential] [Request, 226 lines] D1641: Implement PCI SR-IOV initialization methods

rstone created this revision. rstone added a reviewer: jfvogel. rstone added a subscriber: freebsd-net. REVISION SUMMARY Implement the two new PCI methods used to initialize and uninitialize SR-IOV on a PF. On initialization, the PF must allocate a VEB, which is a virtual switch instance im

[Differential] [Request, 118 lines] D1642: Implement PCI SR-IOV method to initialize individual VFs

rstone created this revision. rstone added a reviewer: jfvogel. rstone added a subscriber: freebsd-net. REVISION SUMMARY For each VF that has been allocated, the PF driver must allocate a Virtual Switch Interface (VSI), which is a virtual port on the virtual Fortville switch (VEB). In d

[Differential] [Request, 225 lines] D1643: Implement resetting a VF

rstone created this revision. rstone added a reviewer: jfvogel. rstone added a subscriber: freebsd-net. REVISION SUMMARY Add the ability to do a full reset of a VF's state. This involves stopping the VF in hardware, releasing all resources, re-allocating a new VSI, and finally allocating th

[Differential] [Request, 1, 330 lines] D1640: Refactor network stack state separate from VSI state

rstone created this revision. rstone added a reviewer: jfvogel. rstone added a subscriber: freebsd-net. REVISION SUMMARY Currently the ixl driver has a struct ixl_vsi that it uses to track both network stack interface state (e.g. the ifnet) and Virtual Station Interface (VSI) state in the ha

[Differential] [Request, 21 lines] D1645: Add support for VERSION VC message

rstone created this revision. rstone added a reviewer: jfvogel. rstone added a subscriber: freebsd-net. REVISION SUMMARY The VC message format is versioned. Implement the VERSION VC message, which the VF uses to discover the version of the VC protocol implemented by the PF. The VF will not

[Differential] [Request, 32 lines] D1647: Add support for GET_VF_RESOURCES VC message

rstone created this revision. rstone added a reviewer: jfvogel. rstone added a subscriber: freebsd-net. REVISION SUMMARY This message is sent by the VF driver to discover the capabilities of the VF, like the number of rx/tx queues supported. REVISION DETAIL https://reviews.freebsd.org/D1647

[Differential] [Request, 152 lines] D1644: Add infrastructure for handling the VC msg channel from VFs

rstone created this revision. rstone added a reviewer: jfvogel. rstone added a subscriber: freebsd-net. REVISION SUMMARY The VF drivers do not have unrestricted access to the hardware. To configure the VF (e.g. set the MAC address), they are required to send a message the PF requesting the c

Re: network locks up with udp traffic

> On 25 янв. 2015 г., at 0:10, Adrian Chadd wrote: > > Ah, try adding STACK to your config? > It's already there. ___ freebsd-net@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to "freebs

[Differential] [Request, 68 lines] D1651: Add support for ENABLE/DISABLE_QUEUES VC messages

rstone created this revision. rstone added a reviewer: jfvogel. rstone added a subscriber: freebsd-net. REVISION SUMMARY These messages are sent by the VF to enable or disable packet reception and transmission from its queues. Despite the fact that a bitmask of queues is provided in the mes

[Differential] [Request, 38 lines] D1648: Add stubs for deprecated VC messages

rstone created this revision. rstone added a reviewer: jfvogel. rstone added a subscriber: freebsd-net. REVISION SUMMARY There are a couple of VC message types defined in the enum that seem to have been deprecated early in the Fortville SR-IOV development process. Explicitly implement them

[Differential] [Request, 19 lines] D1646: Add support for RESET_VF VC message

rstone created this revision. rstone added a reviewer: jfvogel. rstone added a subscriber: freebsd-net. REVISION SUMMARY This VC message is used by the VF driver to request a full reset of the VF state. Unlike all other VC messages, no response is sent to the message. REVISION DETAIL htt

[Differential] [Request, 165 lines] D1650: Add support for CONFIG_IRQ_MAP VC message

rstone created this revision. rstone added a reviewer: jfvogel. rstone added a subscriber: freebsd-net. REVISION SUMMARY This message is used by the VF to request that its RX and TX queues be associated with particular MSI-X vectors. This configuration is performed via a fairly complicated

[Differential] [Request, 161 lines] D1652: Add support for ADD/DEL_ETHER_ADDRESS VC messages

rstone created this revision. rstone added a reviewer: jfvogel. rstone added a subscriber: freebsd-net. REVISION SUMMARY The ADD message is sent by the VF to request that a MAC address filter be added to the Fortville switch to direct packets destined to that MAC be routed to the VF's VSI.

[Differential] [Request, 49 lines] D1654: Add support for CONFIG_PROMISCUOUS_MODE VC message

rstone created this revision. rstone added a reviewer: jfvogel. rstone added a subscriber: freebsd-net. REVISION SUMMARY This message is sent by the VF driver to request that the VSI be placed in or taken out of promiscuous mode. REVISION DETAIL https://reviews.freebsd.org/D1654 AFFECTED F

[Differential] [Request, 177 lines] D1649: Add support for CONFIG_VSI_QUEUES VC message

rstone created this revision. rstone added a reviewer: jfvogel. rstone added a subscriber: freebsd-net. REVISION SUMMARY This message is sent by the VF to request that the PF configure its RX and TX queues. The PF is required to validate that the request is sane before applying the configur

[Differential] [Request, 41 lines] D1655: Add support for GET_STATS VC message

rstone created this revision. rstone added a reviewer: jfvogel. rstone added a subscriber: freebsd-net. REVISION SUMMARY The Fortville hardware tracks a small number of per-VSI counters, like packet and byte counters. The GET_STATS message is sent by the VF driver to request a copy of the c

[Differential] [Request, 121 lines] D1653: Add support for ADD/DEL_VLAN VC messages

rstone created this revision. rstone added a reviewer: jfvogel. rstone added a subscriber: freebsd-net. REVISION SUMMARY These messages are sent by the PF to change configuration related to the hardware vlan filtering and stripping. If a VF is subscribed to one or more vlans then the vlan h

[Differential] [Request, 34 lines] D1657: Allow VFs to run while the PF is admin down

rstone created this revision. rstone added a reviewer: jfvogel. rstone added a subscriber: freebsd-net. REVISION SUMMARY Ensure that the admin queue interrupt stays enabled as long as we have VFs. It is a requirement that the VFs be able to keep running correctly even if the PF ifnet is adm

[Differential] [Request, 65 lines] D1656: Handle VFLR events from VFs

rstone created this revision. rstone added a reviewer: jfvogel. rstone added a subscriber: freebsd-net. REVISION SUMMARY If a VF resets itself via a Function Level Reset (FLR) bit, the PF receives an interrupt with a VFLR event. The PF is required to re-initialize the VF state when this hap

[Differential] [Request, 24 lines] D1659: Register ixl as an SR-IOV-capable driver during attach

rstone created this revision. rstone added a reviewer: jfvogel. rstone added a subscriber: freebsd-net. REVISION SUMMARY Now that the driver's SR-IOV implementation is complete, during attach register the device as being SR-IOV capable. This will cause the SR-IOV infrastructure to create th

[Differential] [Request, 34 lines] D1658: Add sysctls for per-VF hardware counters

rstone created this revision. rstone added a reviewer: jfvogel. rstone added a subscriber: freebsd-net. REVISION SUMMARY For debugging purposes, add sysctls that reflect the value of each VF's hardware counters. REVISION DETAIL https://reviews.freebsd.org/D1658 AFFECTED FILES sys/dev/ixl

[Differential] [Request, 23 lines] D1661: Add some security-related config parameters

rstone created this revision. rstone added a reviewer: jfvogel. rstone added a subscriber: freebsd-net. REVISION SUMMARY Allow the administrator to prevent VFs from sending packets with a source MAC that is not the VF's MAC (anti-spoof). Default this to on. Allow the administrator to g

[Differential] [Request, 21 lines] D1660: Add support for mac-addr parameter

rstone created this revision. rstone added a reviewer: jfvogel. rstone added a subscriber: freebsd-net. REVISION SUMMARY Implement a mac-addr SR-IOV configuration parameter, which the administrator can specify to force a VF to use a particular MAC address. By default, the VF driver cannot c

Re: network locks up with udp traffic

On 24 January 2015 at 14:45, Dmitry Sivachenko wrote: > >> On 25 янв. 2015 г., at 0:10, Adrian Chadd wrote: >> >> Ah, try adding STACK to your config? >> > > > It's already there. Hm, not sure why you're not getting the full stacks in the witness dump. Maybe try KDB and KDB_TRACE ? -adrian __

Re: network locks up with udp traffic

> On 25 янв. 2015 г., at 1:45, Dmitry Sivachenko wrote: > > >> On 25 янв. 2015 г., at 0:10, Adrian Chadd wrote: >> >> Ah, try adding STACK to your config? >> > > > It's already there. Ah, sorry, it is options DDB that was missing. Now we have full stack: lock order reversal: 1st 0xff

Re: network locks up with udp traffic

ok, cool. You definitely have enough for a freebsd bug report. Would you mind adding it? https://bugs.freebsd.org/submit/ -a On 24 January 2015 at 15:07, Dmitry Sivachenko wrote: > >> On 25 янв. 2015 г., at 1:45, Dmitry Sivachenko wrote: >> >> >>> On 25 янв. 2015 г., at 0:10, Adrian Chadd

[Bug 197059] network locks up with IPv6 udp traffic

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=197059 Bug ID: 197059 Summary: network locks up with IPv6 udp traffic Product: Base System Version: 10.0-STABLE Hardware: Any OS: Any Status: New Severity: