Repository navigation
ethernet does not work as it keeps flipping #14
Description
I am using meta-riscv/yocto to build a core-image-base where I also ensuring that we set CONFIG_MICREL_PHY=y but this also does not help in bringing up address on the phy, I can see the interface
but it never gets IP, udhcpc keeps trying. I also disabled udhcpc running on boot to avoid any races. And launch it manually after logging in. But it results in same results
root@beaglev-starlight-jh7100:/sbin# ifup eth0
[ 55.568281] stmmaceth 10020000.gmac eth0: PHY [stmmac-0:07] driver [Micrel KSZ9031 Gigabit PHY] (irq=POLL)
[ 55.580754] dwmac1000: Master AXI performs fixed burst length
[ 55.586626] stmmaceth 10020000.gmac eth0: No Safety Features support found
[ 55.593572] stmmaceth 10020000.gmac eth0: No MAC Management Counters available
[ 55.600790] stmmaceth 10020000.gmac eth0: IEEE 1588-2008 Advanced Timestamp supported
[ 55.609898] stmmaceth 10020000.gmac eth0: configuring for phy/rgmii-txid link mode
udhcpc: started, v1.33.1
udhcpc: sending discover
udhcpc: sending discover
[ 59.752954] stmmaceth 10020000.gmac eth0: Link is Up - 1Gbps/Full - flow control rx/tx
[ 59.760900] IPv6: ADDRCONF(NETDEV_CHANGE): eth0: link becomes ready
udhcpc: sending discover
udhcpc: no lease, forking to background
root@beaglev-starlight-jh7100:/sbin# [ 70.071047] ------------[ cut here ]------------
[ 70.075716] NETDEV WATCHDOG: eth0 (stmmaceth): transmit queue 0 timed out
[ 70.082809] WARNING: CPU: 1 PID: 0 at net/sched/sch_generic.c:442 dev_watchdog+0x226/0x22a
[ 70.091075] Modules linked in:
[ 70.094163] CPU: 1 PID: 0 Comm: swapper/1 Not tainted 5.10.6-yocto-standard-starfive #1
[ 70.102167] epc: ffffffe0006535c4 ra : ffffffe0006535c4 sp : ffffffe07ff0fc20
[ 70.109300] gp : ffffffe000ccab78 tp : ffffffe07fecf500 t0 : ffffffe000c19508
[ 70.116518] t1 : ffffffffffffffff t2 : 0000000000000000 s0 : ffffffe07ff0fc70
[ 70.123735] s1 : 0000000000000000 a0 : 000000000000003d a1 : 0000000000000000
[ 70.130952] a2 : 0000000000000000 a3 : 0000000000000000 a4 : 0000000000000000
[ 70.138172] a5 : 0000000000000000 a6 : 0000000000000000 a7 : 0000000000735049
[ 70.145390] s2 : ffffffe080cd83d0 s3 : ffffffe080cd83a0 s4 : ffffffe000c08a40
[ 70.152611] s5 : ffffffe080cd8000 s6 : ffffffffffffffff s7 : 0000000000000004
[ 70.159830] s8 : ffffffe08016a940 s9 : ffffffe00065339e s10: ffffffe07ff0fca0
[ 70.167048] s11: ffffffe1f8d72600 t3 : 0000000000000002 t4 : 0000000000000402
[ 70.174267] t5 : ffffffe000c84770 t6 : ffffffe07ff0f958
[ 70.179585] status: 0000000200000120 badaddr: ffffffe07fecf508 cause: 0000000000000003
[ 70.187495] ---[ end trace 1526d54b5fd7b7e7 ]---
[ 70.192267] stmmaceth 10020000.gmac eth0: Reset adapter.
[ 70.338374] stmmaceth 10020000.gmac eth0: PHY [stmmac-0:07] driver [Micrel KSZ9031 Gigabit PHY] (irq=POLL)
[ 70.350892] dwmac1000: Master AXI performs fixed burst length
[ 70.356745] stmmaceth 10020000.gmac eth0: No Safety Features support found
[ 70.363677] stmmaceth 10020000.gmac eth0: No MAC Management Counters available
[ 70.370883] stmmaceth 10020000.gmac eth0: IEEE 1588-2008 Advanced Timestamp supported
[ 70.378751] stmmaceth 10020000.gmac eth0: configuring for phy/rgmii-txid link mode
[ 78.712955] stmmaceth 10020000.gmac eth0: Link is Up - 1Gbps/Full - flow control rx/tx
[ 78.720898] IPv6: ADDRCONF(NETDEV_CHANGE): eth0: link becomes ready
I am using starfive_vic7100_evb_sd_net_defconfig but same is seen with fedora defconfig as well and also with https://github.com/starfive-tech/freelight-u-sdk/blob/starfive/conf/sdk_210209_defconfig
Activity
pdp7 commentedon May 12, 2021
@kraj thank you for opening this issue, are you using the Fedora branch in this repo that Wei Fu (@tekkamanninja) created?
Wei Fu told me he is staying on 5.10 as @rwmjones is using that for Fedora/RISC-V however the 5.12 branch from @esmil does seem to be working well for some of the beta developers that have tried it
davidlt commentedon May 12, 2021
I plan to update Fedora/RISCV to v5.12.X as soon as that plans in f33 branch. That usually takes few weeks after new major kernel is released and the current stable kernel goes away. I still see 5.11.19 on kernel.org marked as stable (probably not for long). The next Fedora/RISCV F33 images I will produce will be be v5.12.X based.
kraj commentedon May 12, 2021
I added
stmmac.chain_mode=1to kernel cmdline and that made all the difference. ethernet gets IP now.irqchip/gic-v3: Do not enable irqs when handling spurious interrups
pdp7 commentedon May 18, 2021
@kraj Is the original problem of
NETDEV WATCHDOG: eth0 (stmmaceth): transmit queue 0 timed outstill happening?riscv: mm: Fix W+X mappings at boot
pdp7 commentedon May 27, 2021
@kraj is this still a problem or should I close?
pdp7 commentedon May 27, 2021
Khem confirmed with me on Slack that this is fixed
riscv: mm: Fix W+X mappings at boot
riscv: mm: Fix W+X mappings at boot
perf machine: Initialize srcline string member in add_location struct
31 remaining items
perf ui browser: Don't save pointer to stack memory
perf ui browser: Don't save pointer to stack memory
btrfs: fix crash on racing fsync and size-extending write into prealloc
mm/huge_memory: don't unpoison huge_zero_folio
tracing: Build event generation tests only as modules
mm/huge_memory: don't unpoison huge_zero_folio
tracing: Build event generation tests only as modules
ocfs2: fix DIO failure due to insufficient transaction credits
btrfs: fix crash on racing fsync and size-extending write into prealloc
mm/huge_memory: don't unpoison huge_zero_folio
tracing: Build event generation tests only as modules
ocfs2: fix DIO failure due to insufficient transaction credits
USB: hub: Ignore non-compliant devices with too many configs or inter…
wifi: ath9k: add range check for conn_rsp_epid in htc_connect_service()
perf annotate: Use jump__delete when freeing LoongArch jumps