<Namidairo>
meanwhile I get a whole 50mbit up for a gigabit connection
<stintel>
sounds like Belgium
<nitroshift>
stintel, o/
<nitroshift>
stintel, i'll send you a bucker of internet :p
<nitroshift>
*buker / bucket
<stintel>
I've 1000/600, I'm not complaining
<stintel>
I could get 10000/10000 but with guaranteed 6Mbps that seems a bit silly
rua has quit [Quit: Leaving.]
rua has joined #openwrt-devel
<russell-->
i'm getting an snmpd segfault during start up ([ 50.628844] do_page_fault(): sending SIGSEGV to snmpd for invalid write access to 00000000, but not on restart
<russell-->
any suggestions about how to track down?
<karlp>
huh, I have an snmpd agentx that gets a segfault on _stop_ writing to 0000000, I've kinda lent towards it being a net-snmp bug and have just been ignoring it for a while :|
<russell-->
from the behavior, it feels like some kind of race condition with interfaces appearing and disappearing
floof58 has quit [Ping timeout: 480 seconds]
floof58 has joined #openwrt-devel
pmelange has joined #openwrt-devel
pmelange has quit [Ping timeout: 480 seconds]
<neggles>
Namidairo: australian detected
<neggles>
nitroshift: the ip forwarding offload stuff is indeed present, but it relies on cavium extensions to the kernel
<Namidairo>
will neither confirm or deny
<neggles>
you can't get hw-accel on a MIPS octeon without using the hacky cavium kernel, or porting their hacky code to a new kernel
dedeckeh has quit [Remote host closed the connection]
<neggles>
nitroshift: you can get 4.14.something or use the UBNT ER 4.9.79, and a 5.4 kernel might exist *somewhere* - they've published a toolchain that looks like it's set up to build a 5.4
<neggles>
but... at that point, just use EdgeOS
<nitroshift>
neggles, i'm tired of their gui / lack of updates
<neggles>
that's fair
<neggles>
the CLI is pretty nice, though. it's a shame there's no VyOS for MIPS64
<nitroshift>
neggles, cli is nice indeed
<neggles>
the VyOS team would probably have the resources to sort out the necessary drivers / offload stuff with marvell, they've got newer versions of the same accel engines in the ARM octeons that VyOS supports
<neggles>
octeon does not support zboot and these only have 8MB of NOR - CFI parallel NOR, underneath a heatsink that's glued to the main SoC, so not exactly easy to upgrade
<neggles>
nitroshift: all that aside, PPPoE is almost certainly your limiter - based on the software-only NAT throughput (offload disabled) I can get out of an EdgeRouter Infinity, the quad-core ER-4/6 should be able to do 1G no trouble. These chips have a hardware PPPoE accelerator, iirc, so I guess MIPS is bad at doing PPPoE in software…
knoppiks has quit [Quit: Page closed]
Acinonyx_ has joined #openwrt-devel
Acinonyx has quit [Ping timeout: 480 seconds]
<nitroshift>
neggles, i'll just stick to whatever edgeos version it's running at the moment and don't open its webpage
<nitroshift>
not to mention their ipv6 setup steps are a true nightmare
minimal has joined #openwrt-devel
nitroshift has quit [Quit: Gone that way --->]
pmelange1 has joined #openwrt-devel
pmelange has quit [Read error: Connection reset by peer]
Slimey has joined #openwrt-devel
kenny has joined #openwrt-devel
danitool has quit [Quit: Cubum autem in duos cubos, aut quadratoquadratum in duos quadratoquadratos]
danitool has joined #openwrt-devel
T-Bone has joined #openwrt-devel
f00b4r0 has quit [Ping timeout: 480 seconds]
<hurricos>
neggles: why not just buy NAND and stick it to the back? You just need some solder paste and a heat gun.
<hurricos>
And yes, the right NAND ....
<stintel>
would still require a driver to be ported to the linux kernel
<hurricos>
All of the cards I got had been roughed up pretty bad. I paid $2 apiece, so ... of course :^)
<hurricos>
(after shipping!)
<hurricos>
I wonder if OCTEONII needs a clock for its USB? If not, I wonder where the pads are.
<hurricos>
I also really wonder about the square of pads on the far end of the card. I've always imagined driving these guys with external power
<PaulFertser>
Hm, solder paste is hard to use without a specific stencil or some dispenser.
<hurricos>
paulfertser: Thankfully there's a fully pre-populated TSOP56 pad
<PaulFertser>
hurricos: but how will the solder paste help?
<hurricos>
I just feel there's not ENOUGH solder pre-balled.
<hurricos>
perhaps enough flux might be enough, not sure
<hurricos>
wow, k9k8g08u0d sells for $15 new?!
<hurricos>
that's some expensive flash
<PaulFertser>
In this case some drag soldering (not soldering in drag) should help.
<Habbie>
drag soldering in drag
<Habbie>
just tie down your hair
<hurricos>
unless you like the smell of burning :^)
<PaulFertser>
Whatever makes you more comfy and the hand steady.
<hurricos>
I think I fundamentally misunderstood what NAND is. While you can have SPI NAND, it looks like the NAND on this board is "just a NAND flash interface"
<hurricos>
so no tricky, "I'll just buy a piece of 32MB TSOP48 NOR that's SPI compatible" or anything
<hurricos>
I don't know how any of this works, I feel bad. I should have gone and done EE :(
<hurricos>
Oh, but
<hurricos>
I also understand now how the MR16 can boot using a bootcmd like `bootm 0xbf0a0000`, because NOR can be accessed at random, so it can just map directly into memory
<PaulFertser>
Yes, and it has no bad sectors and no OOB data to store information about that and ECC codes.
<hurricos>
And yet, NOR has lower durability than NAND.
<hurricos>
In terms of write/erase cycles. I guess NOR manufacturers accept lower yields?
<hurricos>
Ahhhhhh, duh. No, the technology is just completely different
<PaulFertser>
hurricos: in a typical MCU you have NOR memory which can survive 10000 erase cycles. I'm not sure regular NAND can handle anywhere near that.
<hurricos>
NAND can handle way more than that per bit, but the technology forces you to access whole blocks at once, and if any one bit breaks down, the whole block or rank or whatever can become unreliable
<hurricos>
because of how the memory itself is addressed, so far as I've just learned
eck has left #openwrt-devel [PIRCH98:WIN 95/98/WIN NT:1.0 (build 1.0.1.1190)]
Borromini has left #openwrt-devel [#openwrt-devel]
pmelange1 has left #openwrt-devel [#openwrt-devel]
<karlp>
man, I'd completely forgotten just how fast openwrt is on an x86 VM, even just one core, with 512MB ram, instead of the little ath79 target I'm normally on!
<karlp>
just blisteringly fast
<stintel>
ath79 is aged ... :)
<karlp>
shush.
<karlp>
I do have the proto of the next gen hardware on my desk, which is sunxi, so sure, I'm upgrading
<karlp>
but had some working frrom home recently, and decided I really needed to up my emulation game, and got x86 qemy working again and it's delightful
<stintel>
yeah, have an x86 vm in ovh as central vpn endpoint, also runs freeswitch, it's really nice
<stintel>
but I also like my new main router - the firebox m300
<neggles>
hurricos: one of my cards has NAND, the other doesn't, but I've ordered two of the same chip that's on the one with NAND to solder onto the other, hopefully it works
<neggles>
and yeah, SPI NAND doesn't require much effort, most of it handles the NAND stuff internally, but raw NAND is much more complicated
<neggles>
the lack of endurance is partly due to having to erase whole pages at once, but it also just doesn't handle being erased as well as NOR does
<neggles>
and yeah, you can't bit-map raw NAND because of the per-page ECC, that's why almost everything will use a small SPI (or parallel/CFI, as is the case on these cards) NOR for the bootloader which then has a full NAND driver
<neggles>
atm we have NOR that works in u-boot, but not in linux - all the UBNT products boot from USB or eMMC and I suspect there's a good reason for that :P
<karlp>
stintel: firebox m300 is that completely eol unobtainium thing right?
<Slimey>
lol
<stintel>
karlp: not really, it's not EOL yet and it's already on ebay
<slh>
just for reference, a(n over a) decade old Atom N270 can do ~650 MBit/s routing, very snappy feeling with OpenWrt, but too power hungry for production usage (~20-22 watts idle, up to 28-29 watts while routing at full speed); Atom 330 can just almost achieve 1 GBit/s (not quite steadily, but >850 MBit/s at all times), but +8 watts