Photon OS

Expand all | Collapse all

MELLANOX OFED Install

  • 1.  MELLANOX OFED Install

    Broadcom Employee
    Posted Dec 06, 2024 09:35 AM

    Hi guys,

    I need to install MELLANOX OFED in photon3, but I cannot find a suitable software package from NVIDIA download. I tried out with the community version but didn't work.

    Could you please point me to where I can download the OFED to be installed in photon OS?

    Thank you 

    Marco



  • 2.  RE: MELLANOX OFED Install

    Broadcom Employee
    Posted Dec 07, 2024 05:34 AM

    We do not officially support it. We had a discussion a year ago with NVidia to support Photon OS 4.0. But it was de prioritized.




  • 3.  RE: MELLANOX OFED Install

    Broadcom Employee
    Posted Dec 07, 2024 07:29 AM
    Unfortunately we have Telco customers running SRIOV on Mellanox and Photon3
    (TKG 2.1 and previous version) in production. For example Singtel with
    Ericsson UPF.




    On Sat, Dec 7, 2024 at 2:36 PM Alexey Makhalov via Broadcom <
    Mail@broadcom.com> wrote:

    > We do not officially support it. We had a discussion a year ago with
    > NVidia to support Photon OS 4.0. But it was de prioritized. -posted to the
    > "Photon OS" community
    > [image: Broadcom] <https: community.broadcom.com>
    > Photon OS
    > <https: community.broadcom.com communities community-home digestviewer?communitykey=a70674e4-ccb6-46a3-ae94-7ecf16c06e24>
    > Post New Message <broadcom-photonos@connectedcommunity.org>
    > Re: MELLANOX OFED Install
    > <https: community.broadcom.com discussion mellanox-ofed-install#bm0a6da5f4-c337-45be-8b1e-2fe8f67eb9e1>
    > Reply to Group
    > <broadcom_photonos_0a6da5f4-c337-45be-8b1e-2fe8f67eb9e1@connectedcommunity.org?subject=re:+mellanox+ofed+install> Reply
    > to Sender
    > <https: community.broadcom.com communities all-discussions postreply?messagekey=0a6da5f4-c337-45be-8b1e-2fe8f67eb9e1&ListKey=39fd2ab1-ab47-42d9-b812-018ed42c5557&SenderKey=2f39633a-1d2b-4441-9bee-018b2ea13496>
    > [image: Alexey Makhalov]
    > <https: community.broadcom.com network members profile?userkey=2f39633a-1d2b-4441-9bee-018b2ea13496>
    > Dec 7, 2024 5:34 AM
    > Alexey Makhalov
    > <https: community.broadcom.com network members profile?userkey=2f39633a-1d2b-4441-9bee-018b2ea13496>
    >
    > We do not officially support it. We had a discussion a year ago with
    > NVidia to support Photon OS 4.0. But it was de prioritized.
    > *Reply to Group Online
    > <https: community.broadcom.com communities all-discussions postreply?messagekey=0a6da5f4-c337-45be-8b1e-2fe8f67eb9e1&ListKey=39fd2ab1-ab47-42d9-b812-018ed42c5557>*
    > *Reply to Group via Email
    > <broadcom_photonos_0a6da5f4-c337-45be-8b1e-2fe8f67eb9e1@connectedcommunity.org?subject=re:+mellanox+ofed+install>*
    > *View Thread
    > <https: community.broadcom.com discussion mellanox-ofed-install#bm0a6da5f4-c337-45be-8b1e-2fe8f67eb9e1>*
    > *Recommend
    > <https: community.broadcom.com:443 discussion mellanox-ofed-install?messagekey=0a6da5f4-c337-45be-8b1e-2fe8f67eb9e1&cmd=rate&cmdarg=add#bm0a6da5f4-c337-45be-8b1e-2fe8f67eb9e1>*
    > *Forward
    > <https: community.broadcom.com communities all-discussions forwardmessages?messagekey=0a6da5f4-c337-45be-8b1e-2fe8f67eb9e1&ListKey=39fd2ab1-ab47-42d9-b812-018ed42c5557>*
    > *Flag as Inappropriate
    > <https: community.broadcom.com discussion mellanox-ofed-install?markappropriate=0a6da5f4-c337-45be-8b1e-2fe8f67eb9e1#bm0a6da5f4-c337-45be-8b1e-2fe8f67eb9e1>*
    >
    > -------------------------------------------
    > Original Message:
    > Sent: Dec 05, 2024 05:43 AM
    > From: Marco Stura
    > Subject: MELLANOX OFED Install
    >
    > Hi guys,
    >
    > I need to install MELLANOX OFED in photon3, but I cannot find a suitable
    > software package from NVIDIA download. I tried out with the community
    > version but didn't work.
    >
    > Could you please point me to where I can download the OFED to be installed
    > in photon OS?
    >
    > Thank you
    >
    > Marco
    >
    >
    >
    >
    > You are receiving this notification because you followed the 'MELLANOX
    > OFED Install' message thread. If you do not wish to follow this, please
    > click here
    > <https: community.broadcom.com higherlogic common unfollow.aspx?userkey=09f7b9fa-7ac0-4e97-bb88-018b2e955c72&sKey=KeyRemoved&ItemKey=dd6c990a-9ec0-4646-b1d4-01939668aa80>.
    >
    >
    > Update your email preferences
    > <https: community.broadcom.com go.aspx?c=Preferences§ion=email> to
    > choose the types of email you receive
    >
    > Unsubscribe from all participation emails
    > <https: community.broadcom.com higherlogic egroups unsubscribe.aspx?userkey=09f7b9fa-7ac0-4e97-bb88-018b2e955c72&sKey=KeyRemoved&mClass=Social>
    >

    --
    This electronic communication and the information and any files transmitted
    with it, or attached to it, are confidential and are intended solely for
    the use of the individual or entity to whom it is addressed and may contain
    information that is confidential, legally privileged, protected by privacy
    laws, or otherwise restricted from disclosure to anyone else. If you are
    not the intended recipient or the person responsible for delivering the
    e-mail to the intended recipient, you are hereby notified that any use,
    copying, distributing, dissemination, forwarding, printing, or copying of
    this e-mail is strictly prohibited. If you received this e-mail in error,
    please return the e-mail to the sender, delete it from your computer, and
    destroy any printed copy of it.




  • 4.  RE: MELLANOX OFED Install

    Posted Dec 08, 2024 03:56 AM

    TKG 2.1 and TKG 2.4 for vSphere comes with Photon OS 3, indeed.
    Accordingly to Release Notes - NVIDIA Docs OFED latest version 24.10-1.1.4.0, December 2024, does not support ConnectX-3 anymore. Which adapter do you use?

    Did a quick test, and Photon OS 3 (4.19.324-1) on Azure, deployed as Standard F4s v2 with accelerated networking enabled, gets a ConnectX4 adapter --> Mellanox Technologies MT27710 Family [ConnectX-4 Lx Virtual Function] [15b3:1016] (rev 80).

    If I understood the install process correctly, first, an ISO is created for the kernel version used, see Installing the Driver - NVIDIA Docs

    Afaik the contrary of "officially support" means mission control your own custom Photon OS image. 




  • 5.  RE: MELLANOX OFED Install

    Broadcom Employee
    Posted Dec 08, 2024 07:45 AM
    I use ConnetctX-6. I managed to install the OFED by tweaking the install.pl
    in the Community version.

    The OFED, rpm version, is compiled and installed in the running OS. No ISO.

    Anyway, Ihit other issues now :-)

    --
    This electronic communication and the information and any files transmitted
    with it, or attached to it, are confidential and are intended solely for
    the use of the individual or entity to whom it is addressed and may contain
    information that is confidential, legally privileged, protected by privacy
    laws, or otherwise restricted from disclosure to anyone else. If you are
    not the intended recipient or the person responsible for delivering the
    e-mail to the intended recipient, you are hereby notified that any use,
    copying, distributing, dissemination, forwarding, printing, or copying of
    this e-mail is strictly prohibited. If you received this e-mail in error,
    please return the e-mail to the sender, delete it from your computer, and
    destroy any printed copy of it.




  • 6.  RE: MELLANOX OFED Install

    Posted Dec 08, 2024 07:58 AM

    What sort of issues? Not sure if we talk about the same. The commmunity version contains the source files. Would you mind to share the steps before the issue(s) and the output logfile?




  • 7.  RE: MELLANOX OFED Install

    Broadcom Employee
    Posted Dec 09, 2024 01:10 AM
    OK, let me give the full background.

    We are using TKG 2.1.1 with BYOD template built by VMware Telco (ex SDE
    now). That template has got mlx5_core and rdma drivers but not loaded at
    boot. I made them loaded at boot and then added SRIOV-VF backed adapters.
    Then I noticed the mlx5_core was not binding automatically, hence no links
    created on those adapters.

    I then created a script to bind the Mellanox SRIOV-VF backed adapters if
    any.

    #!/bin/bash

    # Check for Mellanox devices and fetch their PCI addresses
    echo "Scanning for Mellanox network cards..."
    MLX_CARDS=$(lspci | grep Mellanox | awk '{print $1}')

    # Check if any cards were found
    if [ -z "$MLX_CARDS" ]; then
    echo "No Mellanox network cards detected!"
    exit 1
    fi

    # Count the number of cards detected
    CARD_COUNT=$(echo "$MLX_CARDS" | wc -l)
    echo "Found $CARD_COUNT Mellanox network card(s):"
    echo "$MLX_CARDS"

    # Iterate over each PCI address and bind the mlx5_core driver
    for PCI_ADDR in $MLX_CARDS; do
    echo "Processing Mellanox card at PCI address $PCI_ADDR..."

    # Apply driver override
    echo "mlx5_core" > /sys/bus/pci/devices/0000:$PCI_ADDR/driver_override
    if [ $? -ne 0 ]; then
    echo "Failed to apply driver override for $PCI_ADDR. Skipping..."
    continue
    fi

    # Bind the mlx5_core driver
    echo 0000:$PCI_ADDR > /sys/bus/pci/drivers/mlx5_core/bind
    if [ $? -ne 0 ]; then
    echo "Failed to bind mlx5_core driver for $PCI_ADDR. Skipping..."
    continue
    fi

    echo "Successfully bound mlx5_core driver to $PCI_ADDR."
    done

    echo "Driver binding process completed."


    The script is used by a systemd service that I also created to do it at
    boot. All fine, I can see the mlx5_core binds to the adapters and all looks
    perfect from a PCI and ip link perspective. However, packets are not
    received at the OS IP stack and not sent out from the adapter once the IP
    stack sends e.g. a ping or whatever packet. For example, when I add an IP
    address and ping, tcpdump show the ARP going out from the IP stack but it
    is not sent to the physical network by the adapter.

    Then I thought about replacing the drivers by installing Mellanox OFED
    24.x. Which I did, but I hit the same issue. All looks fine but no packets
    are going through.

    root [ /usr/local/bin ]# lsmod | grep mlx
    mlx5_ib 397312 0
    ib_uverbs 143360 2 rdma_ucm,mlx5_ib
    ib_core 303104 8
    rdma_cm,ib_ipoib,iw_cm,ib_umad,rdma_ucm,ib_uverbs,mlx5_ib,ib_cm
    mlx5_core 1552384 1 mlx5_ib
    mlxfw 24576 1 mlx5_core
    mlxdevm 172032 1 mlx5_core
    auxiliary 16384 2 mlx5_ib,mlx5_core
    mlx_compat 40960 13
    rdma_cm,ib_ipoib,mlxdevm,mlxfw,iw_cm,auxiliary,ib_umad,ib_core,rdma_ucm,ib_uverbs,mlx5_ib,ib_cm,mlx5_core
    hwmon 16384 1 mlx5_core

    root [ /usr/local/bin ]# lspci -kvv
    03:00.0 Ethernet controller: Mellanox Technologies ConnectX Family mlx5Gen
    Virtual Function
    DeviceName: pciPassthru30
    Subsystem: Mellanox Technologies Device 0058
    Physical Slot: 64
    Control: I/O- Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr-
    Stepping- SERR- FastB2B- DisINTx+
    Status: Cap+ 66MHz- UDF- FastB2B- ParErr- DEVSEL=fast >TAbort-
    <tabort-></tabort-><mabort->SERR- <perr-></perr->
    Latency: 0, Cache Line Size: 32 bytes
    NUMA node: 0
    Region 0: Memory at fcd00000 (64-bit, prefetchable) [size=1M]
    Capabilities: [60] Express (v2) Endpoint, MSI 00
    DevCap: MaxPayload 128 bytes, PhantFunc 0, Latency L0s
    <64ns, L1 <1us
    ExtTag- AttnBtn- AttnInd- PwrInd- RBE- FLReset+
    SlotPowerLimit 0.000W
    DevCtl: CorrErr- NonFatalErr- FatalErr- UnsupReq-
    RlxdOrd- ExtTag- PhantFunc- AuxPwr- NoSnoop-
    FLReset-
    MaxPayload 128 bytes, MaxReadReq 128 bytes
    DevSta: CorrErr- NonFatalErr- FatalErr- UnsupReq- AuxPwr-
    TransPend-
    LnkCap: Port #0, Speed 5GT/s, Width x32, ASPM L0s, Exit
    Latency L0s <64ns
    ClockPM- Surprise- LLActRep- BwNot- ASPMOptComp-
    LnkCtl: ASPM Disabled; RCB 64 bytes Disabled- CommClk-
    ExtSynch- ClockPM- AutWidDis- BWInt- AutBWInt-
    LnkSta: Speed 5GT/s (ok), Width x32 (ok)
    TrErr- Train- SlotClk- DLActive- BWMgmt- ABWMgmt-
    DevCap2: Completion Timeout: Not Supported, TimeoutDis-,
    NROPrPrP-, LTR-
    10BitTagComp-, 10BitTagReq-, OBFF Not Supported,
    ExtFmt-, EETLPPrefix-
    EmergencyPowerReduction Not Supported,
    EmergencyPowerReductionInit-
    FRS-, TPHComp-, ExtTPHComp-
    AtomicOpsCap: 32bit- 64bit- 128bitCAS-
    DevCtl2: Completion Timeout: 50us to 50ms, TimeoutDis-,
    LTR-, OBFF Disabled
    AtomicOpsCtl: ReqEn-
    LnkCtl2: Target Link Speed: 2.5GT/s, EnterCompliance-
    SpeedDis-
    Transmit Margin: Normal Operating Range,
    EnterModifiedCompliance- ComplianceSOS-
    Compliance De-emphasis: -6dB
    LnkSta2: Current De-emphasis Level: -6dB,
    EqualizationComplete-, EqualizationPhase1-
    EqualizationPhase2-, EqualizationPhase3-,
    LinkEqualizationRequest-
    Capabilities: [9c] MSI-X: Enable+ Count=12 Masked-
    Vector table: BAR=0 offset=00002000
    PBA: BAR=0 offset=00003000
    Capabilities: [100 v1] Vendor Specific Information: ID=0000 Rev=0
    Len=00c
    Capabilities: [150 v1] Alternative Routing-ID Interpretation (ARI)
    ARICap: MFVC- ACS-, Next Function: 0
    ARICtl: MFVC- ACS-, Function Group: 0
    Kernel driver in use: mlx5_core

    03:01.0 Ethernet controller: Mellanox Technologies ConnectX Family mlx5Gen
    Virtual Function
    DeviceName: pciPassthru31
    Subsystem: Mellanox Technologies Device 0058
    Physical Slot: 65
    Control: I/O- Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr-
    Stepping- SERR- FastB2B- DisINTx+
    Status: Cap+ 66MHz- UDF- FastB2B- ParErr- DEVSEL=fast >TAbort-
    <tabort-></tabort-><mabort->SERR- <perr-></perr->
    Latency: 0, Cache Line Size: 32 bytes
    NUMA node: 0
    Region 0: Memory at fcc00000 (64-bit, prefetchable) [size=1M]
    Capabilities: [60] Express (v2) Endpoint, MSI 00
    DevCap: MaxPayload 128 bytes, PhantFunc 0, Latency L0s
    <64ns, L1 <1us
    ExtTag- AttnBtn- AttnInd- PwrInd- RBE- FLReset+
    SlotPowerLimit 0.000W
    DevCtl: CorrErr- NonFatalErr- FatalErr- UnsupReq-
    RlxdOrd- ExtTag- PhantFunc- AuxPwr- NoSnoop-
    FLReset-
    MaxPayload 128 bytes, MaxReadReq 128 bytes
    DevSta: CorrErr- NonFatalErr- FatalErr- UnsupReq- AuxPwr-
    TransPend-
    LnkCap: Port #0, Speed 5GT/s, Width x32, ASPM L0s, Exit
    Latency L0s <64ns
    ClockPM- Surprise- LLActRep- BwNot- ASPMOptComp-
    LnkCtl: ASPM Disabled; RCB 64 bytes Disabled- CommClk-
    ExtSynch- ClockPM- AutWidDis- BWInt- AutBWInt-
    LnkSta: Speed 5GT/s (ok), Width x32 (ok)
    TrErr- Train- SlotClk- DLActive- BWMgmt- ABWMgmt-
    DevCap2: Completion Timeout: Not Supported, TimeoutDis-,
    NROPrPrP-, LTR-
    10BitTagComp-, 10BitTagReq-, OBFF Not Supported,
    ExtFmt-, EETLPPrefix-
    EmergencyPowerReduction Not Supported,
    EmergencyPowerReductionInit-
    FRS-, TPHComp-, ExtTPHComp-
    AtomicOpsCap: 32bit- 64bit- 128bitCAS-
    DevCtl2: Completion Timeout: 50us to 50ms, TimeoutDis-,
    LTR-, OBFF Disabled
    AtomicOpsCtl: ReqEn-
    LnkSta2: Current De-emphasis Level: -6dB,
    EqualizationComplete-, EqualizationPhase1-
    EqualizationPhase2-, EqualizationPhase3-,
    LinkEqualizationRequest-
    Capabilities: [9c] MSI-X: Enable+ Count=12 Masked-
    Vector table: BAR=0 offset=00002000
    PBA: BAR=0 offset=00003000
    Capabilities: [100 v1] Vendor Specific Information: ID=0000 Rev=0
    Len=00c
    Capabilities: [150 v1] Alternative Routing-ID Interpretation (ARI)
    ARICap: MFVC- ACS-, Next Function: 0
    ARICtl: MFVC- ACS-, Function Group: 0
    Kernel driver in use: mlx5_core

    root [ /usr/local/bin ]# rdma link show
    0/1: mlx5_0/1: state ACTIVE physical_state LINK_UP netdev Net3
    1/1: mlx5_1/1: state ACTIVE physical_state LINK_UP netdev Net4

    root [ /usr/local/bin ]# ibv_devinfo
    hca_id: mlx5_0
    transport: InfiniBand (0)
    fw_ver: 22.38.1002
    node_guid: 0050:56ff:fea4:841f
    sys_image_guid: a088:c203:003b:cd0a
    vendor_id: 0x02c9
    vendor_part_id: 4126
    hw_ver: 0x0
    board_id: DEL0000000027
    phys_port_cnt: 1
    port: 1
    state: PORT_ACTIVE (4)
    max_mtu: 4096 (5)
    active_mtu: 1024 (3)
    sm_lid: 0
    port_lid: 0
    port_lmc: 0x00
    link_layer: Ethernet

    hca_id: mlx5_1
    transport: InfiniBand (0)
    fw_ver: 22.38.1002
    node_guid: 0050:56ff:fea4:3c15
    sys_image_guid: a088:c203:003b:cd0a
    vendor_id: 0x02c9
    vendor_part_id: 4126
    hw_ver: 0x0
    board_id: DEL0000000027
    phys_port_cnt: 1
    port: 1
    state: PORT_ACTIVE (4)
    max_mtu: 4096 (5)
    active_mtu: 1024 (3)
    sm_lid: 0
    port_lid: 0
    port_lmc: 0x00
    link_layer: Ethernet

    If I use TKG 2.5 (Photon5) all works fine without doing anything. The only
    difference I can notice is that in Photon5 there is a PCI controller module
    that is not present in TKG 2.1.1 Photon3, pci_hyperv_intf that has to do
    with HYPERV but it may help with something???

    root [ /home/capv ]# lsmod | grep mlx
    mlx5_ib 368640 0
    ib_uverbs 151552 1 mlx5_ib
    ib_core 385024 2 ib_uverbs,mlx5_ib
    mlx5_core 1433600 1 mlx5_ib
    pci_hyperv_intf 16384 1 mlx5_core

    Do you understand my issue now?

    Br,
    Marco

    --
    This electronic communication and the information and any files transmitted
    with it, or attached to it, are confidential and are intended solely for
    the use of the individual or entity to whom it is addressed and may contain
    information that is confidential, legally privileged, protected by privacy
    laws, or otherwise restricted from disclosure to anyone else. If you are
    not the intended recipient or the person responsible for delivering the
    e-mail to the intended recipient, you are hereby notified that any use,
    copying, distributing, dissemination, forwarding, printing, or copying of
    this e-mail is strictly prohibited. If you received this e-mail in error,
    please return the e-mail to the sender, delete it from your computer, and
    destroy any printed copy of it.




  • 8.  RE: MELLANOX OFED Install

    Broadcom Employee
    Posted Dec 09, 2024 05:04 AM
    If I look at what pci_hyperv_intf is doing

    * This driver acts as a paravirtual front-end for PCI Express root buses.
    * When a PCI Express function (either an entire device or an SR-IOV
    * Virtual Function) is being passed through to the VM, this driver exposes
    * a new bus to the guest VM. This is modeled as a root PCI bus because
    * no bridges are being exposed to the VM. In fact, with a "Generation 2"
    * VM within Hyper-V, there may seem to be no PCI bus at all in the VM
    * until a device has been exposed using this driver.

    May be it plays a role also in VMware environment. Indeed if I ****
    the module from loading and then I add Mellanox SRIOV-VF adapter, the
    adapter is not detected and mlx5_core is not loading.

    Is there someone in this list that knows exactly what happens with Mellanox
    SRIOV-VF if we don't have this module?

    How to compile Photon3 with the pci_hyperv_intf module?
    If I use rpm2cpio
    https://packages.vmware.com/photon/4.0/photon_srpms_3.0_x86_64/linux-4.19.315-1.ph3.src.rpm
    | cpio -idm. and then I compile and check, there is no pci_hyperv_intf.c
    available, but now I remember in TKG < 2.1.1 I could see that controller
    driver and all was working.


    Br,
    Marco

    On Mon, Dec 9, 2024 at 10:09 AM Marco Stura <marco.stura@broadcom.com>
    wrote:

    > OK, let me give the full background.
    >
    > We are using TKG 2.1.1 with BYOD template built by VMware Telco (ex SDE
    > now). That template has got mlx5_core and rdma drivers but not loaded at
    > boot. I made them loaded at boot and then added SRIOV-VF backed adapters.
    > Then I noticed the mlx5_core was not binding automatically, hence no links
    > created on those adapters.
    >
    > I then created a script to bind the Mellanox SRIOV-VF backed adapters if
    > any.
    >
    > #!/bin/bash
    >
    > # Check for Mellanox devices and fetch their PCI addresses
    > echo "Scanning for Mellanox network cards..."
    > MLX_CARDS=$(lspci | grep Mellanox | awk '{print $1}')
    >
    > # Check if any cards were found
    > if [ -z "$MLX_CARDS" ]; then
    > echo "No Mellanox network cards detected!"
    > exit 1
    > fi
    >
    > # Count the number of cards detected
    > CARD_COUNT=$(echo "$MLX_CARDS" | wc -l)
    > echo "Found $CARD_COUNT Mellanox network card(s):"
    > echo "$MLX_CARDS"
    >
    > # Iterate over each PCI address and bind the mlx5_core driver
    > for PCI_ADDR in $MLX_CARDS; do
    > echo "Processing Mellanox card at PCI address $PCI_ADDR..."
    >
    > # Apply driver override
    > echo "mlx5_core" > /sys/bus/pci/devices/0000:$PCI_ADDR/driver_override
    > if [ $? -ne 0 ]; then
    > echo "Failed to apply driver override for $PCI_ADDR. Skipping..."
    > continue
    > fi
    >
    > # Bind the mlx5_core driver
    > echo 0000:$PCI_ADDR > /sys/bus/pci/drivers/mlx5_core/bind
    > if [ $? -ne 0 ]; then
    > echo "Failed to bind mlx5_core driver for $PCI_ADDR. Skipping..."
    > continue
    > fi
    >
    > echo "Successfully bound mlx5_core driver to $PCI_ADDR."
    > done
    >
    > echo "Driver binding process completed."
    >
    >
    > The script is used by a systemd service that I also created to do it at
    > boot. All fine, I can see the mlx5_core binds to the adapters and all looks
    > perfect from a PCI and ip link perspective. However, packets are not
    > received at the OS IP stack and not sent out from the adapter once the IP
    > stack sends e.g. a ping or whatever packet. For example, when I add an IP
    > address and ping, tcpdump show the ARP going out from the IP stack but it
    > is not sent to the physical network by the adapter.
    >
    > Then I thought about replacing the drivers by installing Mellanox OFED
    > 24.x. Which I did, but I hit the same issue. All looks fine but no packets
    > are going through.
    >
    > root [ /usr/local/bin ]# lsmod | grep mlx
    > mlx5_ib 397312 0
    > ib_uverbs 143360 2 rdma_ucm,mlx5_ib
    > ib_core 303104 8
    > rdma_cm,ib_ipoib,iw_cm,ib_umad,rdma_ucm,ib_uverbs,mlx5_ib,ib_cm
    > mlx5_core 1552384 1 mlx5_ib
    > mlxfw 24576 1 mlx5_core
    > mlxdevm 172032 1 mlx5_core
    > auxiliary 16384 2 mlx5_ib,mlx5_core
    > mlx_compat 40960 13
    > rdma_cm,ib_ipoib,mlxdevm,mlxfw,iw_cm,auxiliary,ib_umad,ib_core,rdma_ucm,ib_uverbs,mlx5_ib,ib_cm,mlx5_core
    > hwmon 16384 1 mlx5_core
    >
    > root [ /usr/local/bin ]# lspci -kvv
    > 03:00.0 Ethernet controller: Mellanox Technologies ConnectX Family mlx5Gen
    > Virtual Function
    > DeviceName: pciPassthru30
    > Subsystem: Mellanox Technologies Device 0058
    > Physical Slot: 64
    > Control: I/O- Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop-
    > ParErr- Stepping- SERR- FastB2B- DisINTx+
    > Status: Cap+ 66MHz- UDF- FastB2B- ParErr- DEVSEL=fast >TAbort-
    > <tabort-></tabort-><mabort->SERR- <perr-></perr->
    > Latency: 0, Cache Line Size: 32 bytes
    > NUMA node: 0
    > Region 0: Memory at fcd00000 (64-bit, prefetchable) [size=1M]
    > Capabilities: [60] Express (v2) Endpoint, MSI 00
    > DevCap: MaxPayload 128 bytes, PhantFunc 0, Latency L0s
    > <64ns, L1 <1us
    > ExtTag- AttnBtn- AttnInd- PwrInd- RBE- FLReset+
    > SlotPowerLimit 0.000W
    > DevCtl: CorrErr- NonFatalErr- FatalErr- UnsupReq-
    > RlxdOrd- ExtTag- PhantFunc- AuxPwr- NoSnoop-
    > FLReset-
    > MaxPayload 128 bytes, MaxReadReq 128 bytes
    > DevSta: CorrErr- NonFatalErr- FatalErr- UnsupReq- AuxPwr-
    > TransPend-
    > LnkCap: Port #0, Speed 5GT/s, Width x32, ASPM L0s, Exit
    > Latency L0s <64ns
    > ClockPM- Surprise- LLActRep- BwNot- ASPMOptComp-
    > LnkCtl: ASPM Disabled; RCB 64 bytes Disabled- CommClk-
    > ExtSynch- ClockPM- AutWidDis- BWInt- AutBWInt-
    > LnkSta: Speed 5GT/s (ok), Width x32 (ok)
    > TrErr- Train- SlotClk- DLActive- BWMgmt- ABWMgmt-
    > DevCap2: Completion Timeout: Not Supported, TimeoutDis-,
    > NROPrPrP-, LTR-
    > 10BitTagComp-, 10BitTagReq-, OBFF Not Supported,
    > ExtFmt-, EETLPPrefix-
    > EmergencyPowerReduction Not Supported,
    > EmergencyPowerReductionInit-
    > FRS-, TPHComp-, ExtTPHComp-
    > AtomicOpsCap: 32bit- 64bit- 128bitCAS-
    > DevCtl2: Completion Timeout: 50us to 50ms, TimeoutDis-,
    > LTR-, OBFF Disabled
    > AtomicOpsCtl: ReqEn-
    > LnkCtl2: Target Link Speed: 2.5GT/s, EnterCompliance-
    > SpeedDis-
    > Transmit Margin: Normal Operating Range,
    > EnterModifiedCompliance- ComplianceSOS-
    > Compliance De-emphasis: -6dB
    > LnkSta2: Current De-emphasis Level: -6dB,
    > EqualizationComplete-, EqualizationPhase1-
    > EqualizationPhase2-, EqualizationPhase3-,
    > LinkEqualizationRequest-
    > Capabilities: [9c] MSI-X: Enable+ Count=12 Masked-
    > Vector table: BAR=0 offset=00002000
    > PBA: BAR=0 offset=00003000
    > Capabilities: [100 v1] Vendor Specific Information: ID=0000 Rev=0
    > Len=00c
    > Capabilities: [150 v1] Alternative Routing-ID Interpretation (ARI)
    > ARICap: MFVC- ACS-, Next Function: 0
    > ARICtl: MFVC- ACS-, Function Group: 0
    > Kernel driver in use: mlx5_core
    >
    > 03:01.0 Ethernet controller: Mellanox Technologies ConnectX Family mlx5Gen
    > Virtual Function
    > DeviceName: pciPassthru31
    > Subsystem: Mellanox Technologies Device 0058
    > Physical Slot: 65
    > Control: I/O- Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop-
    > ParErr- Stepping- SERR- FastB2B- DisINTx+
    > Status: Cap+ 66MHz- UDF- FastB2B- ParErr- DEVSEL=fast >TAbort-
    > <tabort-></tabort-><mabort->SERR- <perr-></perr->
    > Latency: 0, Cache Line Size: 32 bytes
    > NUMA node: 0
    > Region 0: Memory at fcc00000 (64-bit, prefetchable) [size=1M]
    > Capabilities: [60] Express (v2) Endpoint, MSI 00
    > DevCap: MaxPayload 128 bytes, PhantFunc 0, Latency L0s
    > <64ns, L1 <1us
    > ExtTag- AttnBtn- AttnInd- PwrInd- RBE- FLReset+
    > SlotPowerLimit 0.000W
    > DevCtl: CorrErr- NonFatalErr- FatalErr- UnsupReq-
    > RlxdOrd- ExtTag- PhantFunc- AuxPwr- NoSnoop-
    > FLReset-
    > MaxPayload 128 bytes, MaxReadReq 128 bytes
    > DevSta: CorrErr- NonFatalErr- FatalErr- UnsupReq- AuxPwr-
    > TransPend-
    > LnkCap: Port #0, Speed 5GT/s, Width x32, ASPM L0s, Exit
    > Latency L0s <64ns
    > ClockPM- Surprise- LLActRep- BwNot- ASPMOptComp-
    > LnkCtl: ASPM Disabled; RCB 64 bytes Disabled- CommClk-
    > ExtSynch- ClockPM- AutWidDis- BWInt- AutBWInt-
    > LnkSta: Speed 5GT/s (ok), Width x32 (ok)
    > TrErr- Train- SlotClk- DLActive- BWMgmt- ABWMgmt-
    > DevCap2: Completion Timeout: Not Supported, TimeoutDis-,
    > NROPrPrP-, LTR-
    > 10BitTagComp-, 10BitTagReq-, OBFF Not Supported,
    > ExtFmt-, EETLPPrefix-
    > EmergencyPowerReduction Not Supported,
    > EmergencyPowerReductionInit-
    > FRS-, TPHComp-, ExtTPHComp-
    > AtomicOpsCap: 32bit- 64bit- 128bitCAS-
    > DevCtl2: Completion Timeout: 50us to 50ms, TimeoutDis-,
    > LTR-, OBFF Disabled
    > AtomicOpsCtl: ReqEn-
    > LnkSta2: Current De-emphasis Level: -6dB,
    > EqualizationComplete-, EqualizationPhase1-
    > EqualizationPhase2-, EqualizationPhase3-,
    > LinkEqualizationRequest-
    > Capabilities: [9c] MSI-X: Enable+ Count=12 Masked-
    > Vector table: BAR=0 offset=00002000
    > PBA: BAR=0 offset=00003000
    > Capabilities: [100 v1] Vendor Specific Information: ID=0000 Rev=0
    > Len=00c
    > Capabilities: [150 v1] Alternative Routing-ID Interpretation (ARI)
    > ARICap: MFVC- ACS-, Next Function: 0
    > ARICtl: MFVC- ACS-, Function Group: 0
    > Kernel driver in use: mlx5_core
    >
    > root [ /usr/local/bin ]# rdma link show
    > 0/1: mlx5_0/1: state ACTIVE physical_state LINK_UP netdev Net3
    > 1/1: mlx5_1/1: state ACTIVE physical_state LINK_UP netdev Net4
    >
    > root [ /usr/local/bin ]# ibv_devinfo
    > hca_id: mlx5_0
    > transport: InfiniBand (0)
    > fw_ver: 22.38.1002
    > node_guid: 0050:56ff:fea4:841f
    > sys_image_guid: a088:c203:003b:cd0a
    > vendor_id: 0x02c9
    > vendor_part_id: 4126
    > hw_ver: 0x0
    > board_id: DEL0000000027
    > phys_port_cnt: 1
    > port: 1
    > state: PORT_ACTIVE (4)
    > max_mtu: 4096 (5)
    > active_mtu: 1024 (3)
    > sm_lid: 0
    > port_lid: 0
    > port_lmc: 0x00
    > link_layer: Ethernet
    >
    > hca_id: mlx5_1
    > transport: InfiniBand (0)
    > fw_ver: 22.38.1002
    > node_guid: 0050:56ff:fea4:3c15
    > sys_image_guid: a088:c203:003b:cd0a
    > vendor_id: 0x02c9
    > vendor_part_id: 4126
    > hw_ver: 0x0
    > board_id: DEL0000000027
    > phys_port_cnt: 1
    > port: 1
    > state: PORT_ACTIVE (4)
    > max_mtu: 4096 (5)
    > active_mtu: 1024 (3)
    > sm_lid: 0
    > port_lid: 0
    > port_lmc: 0x00
    > link_layer: Ethernet
    >
    > If I use TKG 2.5 (Photon5) all works fine without doing anything. The only
    > difference I can notice is that in Photon5 there is a PCI controller module
    > that is not present in TKG 2.1.1 Photon3, pci_hyperv_intf that has to do
    > with HYPERV but it may help with something???
    >
    > root [ /home/capv ]# lsmod | grep mlx
    > mlx5_ib 368640 0
    > ib_uverbs 151552 1 mlx5_ib
    > ib_core 385024 2 ib_uverbs,mlx5_ib
    > mlx5_core 1433600 1 mlx5_ib
    > pci_hyperv_intf 16384 1 mlx5_core
    >
    > Do you understand my issue now?
    >
    > Br,
    > Marco
    >
    >

    --
    This electronic communication and the information and any files transmitted
    with it, or attached to it, are confidential and are intended solely for
    the use of the individual or entity to whom it is addressed and may contain
    information that is confidential, legally privileged, protected by privacy
    laws, or otherwise restricted from disclosure to anyone else. If you are
    not the intended recipient or the person responsible for delivering the
    e-mail to the intended recipient, you are hereby notified that any use,
    copying, distributing, dissemination, forwarding, printing, or copying of
    this e-mail is strictly prohibited. If you received this e-mail in error,
    please return the e-mail to the sender, delete it from your computer, and
    destroy any printed copy of it.




  • 9.  RE: MELLANOX OFED Install

    Posted Dec 09, 2024 05:30 AM
    Edited by Daniel Casota Dec 09, 2024 05:30 AM

    The test of Photon OS 3 on an Azure virtual machine V2 with accelerated networking simply was to "see" the Mellanox card from the underlying infrastructure as I do not have the equipment on the VMware vSphere homelab side. There is no further value for your case.

    Mellanox Linux drivers loaded at boot means make the kernel to do so. You have to recompile with the appropriate driver settings. In Kconfig there are flags e.g. CONFIG_MLX5_CORE. See Mellanox ConnectX(R) mlx5 core VPI Network Driver - The Linux Kernel documentation . I don't think there is another way beside compile drivers from source, make build your own Photon OS 3 and deploy and maintain Photon OS from that custom iso/ova. This is an embedded software engineer job.

    @Alexey Makhalov, please correct if I'm wrong.




  • 10.  RE: MELLANOX OFED Install

    Broadcom Employee
    Posted Dec 09, 2024 05:50 AM
    Who said on Azure :-)

    I'm running everything onprem VMware platform. FYI, as long as the driver
    module is available in the kernel you can load art boot just by adding a
    .conf in /etc/modules-load.d.
    For example, mlx5_load.conf with content mlx5_core.

    Photon3 has got mlx5_core and rdma modules as well as infiniband verbs, but
    as I explained in details I face another issue that I think related to the
    missing module I mentioned and for which I asked questions to some deep
    expert in this group, I hope I will get an answer sooner or later ...

    Br,
    Marco

    --
    This electronic communication and the information and any files transmitted
    with it, or attached to it, are confidential and are intended solely for
    the use of the individual or entity to whom it is addressed and may contain
    information that is confidential, legally privileged, protected by privacy
    laws, or otherwise restricted from disclosure to anyone else. If you are
    not the intended recipient or the person responsible for delivering the
    e-mail to the intended recipient, you are hereby notified that any use,
    copying, distributing, dissemination, forwarding, printing, or copying of
    this e-mail is strictly prohibited. If you received this e-mail in error,
    please return the e-mail to the sender, delete it from your computer, and
    destroy any printed copy of it.




  • 11.  RE: MELLANOX OFED Install

    Posted Dec 09, 2024 06:29 AM

    If I understood you right, you already tested TKG 2.5 with Photon OS 5 and your use case on same bios and adapters worked flawlessly, right?

    Have you seen this kb SR-IOV: VF "Link Down" is not detected when PF is down for Mellanox ConnectX-5 ? Nvidia recently published quite a few ConnectX-6 fw updates. Did you check the firmware release notes?

    What business value do you expect using TKG 2.1 with a custom Photon OS 3 ?