srslypascal has quit [Remote host closed the connection]
srslypascal has joined #openwrt-devel
ecloud_ has joined #openwrt-devel
wentasah has quit [Remote host closed the connection]
<neggles>
hurricos: oh yeah i have like four LB6Ms
<neggles>
the problem is they pull about 125w before you put any optics in them
<neggles>
once you cross-flash them, they become a TurboIron-24X
<neggles>
they don't use u-boot, and they don't run linux by default
<neggles>
as for support for DSA, no, you won't get that because it's nowhere near remotely sophisticated enough to cover what these ASICs are capable of
<neggles>
however what you *do* get is SAI, Switch Abstraction Interface, as used by Azure SONiC - and here's the BCM source https://github.com/Azure/saibcm-modules
<neggles>
...wrong repo
<neggles>
wait no right repo it's just all in branches
<neggles>
that's for stratadnx, the other stuff is in the sonic-buildimage repo, platform/broadcom
<hurricos>
neggles: Thanks for the information. All good to know. I'm much more used to using big, fast switches together with routers-on-a-stick, so this is all surprising to me
<neggles>
hurricos: oh you're not aware of how much power these ASICs pull?
<neggles>
lemme tell ya
<neggles>
125w is *relatively* low power
<neggles>
things with 48x10G-T ports on them are the worst offenders but there are cisco 1U 48-port switches that eat *four hundred watts at idle*
<neggles>
the way the ASICs are built, an awful lot of them are constant-power regardless of packet flow; idle to full load variance <10W. over half the die area is TCAM for L3 routing at wirespeed
cbeznea has quit [Read error: Connection reset by peer]
<hurricos>
T ... content addressable memory, but what's the T for?
<hurricos>
Ternary. Found the article about it from Cisco. Hmm.
rua has quit [Ping timeout: 480 seconds]
rua has joined #openwrt-devel
<hauke>
neggles: that's 12W per port, that is a lot. I thought the Aquantia 10G PHY with 3.3W was already a lot
cmonroe has quit [Ping timeout: 480 seconds]
cmonroe has joined #openwrt-devel
<neggles>
hauke: this is without PHYs
<neggles>
some 10G-T switches exceed 800W
<neggles>
hurricos: it’s special SRAM that can do a lookup through its entire memory space in one clock cycle
Misanthropos has joined #openwrt-devel
bluew has quit [Quit: Leaving]
<neggles>
IPv4 TCAM stores a prefix, netmask, and nexthop per route, ordered by length of netmask (which indicates which bits are set to “don’t care”). present an IP address on the “address” lines, cycle the clock, and the data lines will contain the nexthop for the first matching entry
<neggles>
longest prefix first routing fully implemented in hardware :D
rua has quit [Ping timeout: 480 seconds]
jlsalvador has joined #openwrt-devel
rua has joined #openwrt-devel
<lynxis>
svanheule: do you have an idea what's the difference between hisgmii and base2500-x on realteks 2.5gbit phys (e.g. 8226 or 8221)?
<robimarko>
That should be the same
<robimarko>
HISGMII is vendor for 2.5G SGMII
<robimarko>
Linux just uses 2500Base-X for that
rua has quit [Ping timeout: 480 seconds]
<lynxis>
i found a gpl'ed driver for rtl8226 phy and they have different values for hisgmii and 2500base-x. I wonder if it's the autoneg on the sgmii link.
mattytap__ has joined #openwrt-devel
mattytap_ has quit [Read error: Connection reset by peer]
<blogic>
robimarko: I check the docs avail to me per OFPROM ... found nothing ...
<blogic>
*check'ed
<blogic>
does sfp 11.5 contain that info inside the src ?
<blogic>
and could you point me at the weirdness that you see ?
rua has joined #openwrt-devel
mattytap__ has quit [Ping timeout: 480 seconds]
Misanthropos has quit [Ping timeout: 480 seconds]
hanetzer1 has quit [Ping timeout: 480 seconds]
mattytap__ has joined #openwrt-devel
mattytap_ has quit [Ping timeout: 480 seconds]
danitool has quit [Read error: Connection reset by peer]
rua has quit [Ping timeout: 480 seconds]
rua has joined #openwrt-devel
<robimarko>
blogic: It was my fault in the end
<robimarko>
QFPROM driver in the kernel works in the end
<robimarko>
I was misunderstanding the reg property for fuses
<robimarko>
CPR region can be figured out from QSDK kernel, thats what I did
<robimarko>
lynxis: Its probably autoneg, Linux just relies on in-band management property for that
danitool has joined #openwrt-devel
mattytap_ has joined #openwrt-devel
mattytap__ has quit [Ping timeout: 480 seconds]
cbeznea1 has quit [Remote host closed the connection]
hanetzer has joined #openwrt-devel
<svanheule>
lynxis: I'm pretty sure robimarko explained it better than anything I could've come up with. I don't have much experience with the networking part of the realtek devices yet
cbeznea has joined #openwrt-devel
<robimarko>
Its not Realtek specific, at least it should not be
<robimarko>
Its kind of an industry standard, unless Realtek managed to mess that as well
<Habbie>
hurricos, zeroes before 0xbf000000 in uboot md, but no hangs :)
<Habbie>
hurricos, uboot strings, kernel image headers, etc. nicely matching up with the mtd table from the boot log
<Habbie>
hurricos, so 0xbf000000 looks right
floof58 has quit [Remote host closed the connection]
mrkiko has joined #openwrt-devel
rua has quit [Ping timeout: 480 seconds]
floof58 has joined #openwrt-devel
rua has joined #openwrt-devel
<russell-->
anybody here have a ubiquiti er-x?
<lynxis>
russell--: yes
<russell-->
can you run "find /rom -type f | xargs sha256sum" and let me know if it throws any errors in dmesg?
<russell-->
and, also, what version it is running
<russell-->
and the approximate amount of time it's been running since the last flash
<mrkiko>
Did someone in the ML notice my comment regarding GL-B2200 PCi wi-fi device being broken? May someone help me fix it up?
<lynxis>
russell--: eta 60 min. the device isn't running and I don't know how long it ran. i would guess a couple days.
<russell-->
lynxis: fwiw, i have seen some squashfs corruption in some long running, i'm trying to discover how widespread it is
<mrkiko>
russell--: the er-x is a mt7621 device, right?
<russell-->
mrkiko: yes
<mrkiko>
rustI saw a strange thing happening on a R6220 I booted after kinda 2 years or something (approx. 1 year); sqaushfs unreadable and errors in the dmesg; sysupgrading did magically fix all the issues
<russell-->
mrkiko: that sounds pretty much the same
<mrkiko>
russell--: so, you're telling me I will experience the corruption again next time I bring it up?
<mrkiko>
I didn't see it across reboots
<Habbie>
hurricos, ah, i learned that xxd can directly consume 'md' output, as long as i tell it about the offset ;)
<mrkiko>
if I'm not mistaken
<russell-->
you can check it now, by scanning all the files in the squashfs. sometimes the corruptions affect reboots, sometimes not
<russell-->
if you reflash before rebooting, then "magic fix"
<russell-->
also, i'm having a weird feeling of dejavu, i'm going to dredge through irc logs
<mrkiko>
russell--: let me know if you find something, Iìm curious
<Habbie>
hurricos, hah, and after 0x400000 bytes, content just repeats :D
<russell-->
2020-11-17 04:04 rsalvaterra: [ 0.631956] nand: WARNING: mt7621-nand: the ECC used on your system is too weak compared to the one required by the NAND chip
shibboleth has quit [Quit: shibboleth]
arthur_melo has quit []
<rsalvaterra>
russell--: Should be safe to ignore. On which kernel are you seeing this? And what device?
<rsalvaterra>
I've seen the warning on my Redmi AC2100, on 5.4, I think.
Tapper has quit [Ping timeout: 480 seconds]
<rsalvaterra>
But the ECC strength is 4 bits, as it should be.
rua has quit [Ping timeout: 480 seconds]
rua has joined #openwrt-devel
<lynxis>
russell--: it seems I don't have one at home. sorry.