[29 Jun 16:00] I mean, I will have time but I can't tell you when [29 Jun 17:31] little question: how often is "hammer2 bulkfree" run in v5.8.1.5.g7c1fb-RELEASE ? [29 Jun 17:31] once every hour? [29 Jun 17:34] sickness I think it's once everyday during the night [29 Jun 17:36] ah, ok, tnx :) [29 Jun 17:36] it's from /etc/periodic/daily/161.clean-hammer2 [29 Jun 17:48] tnx [29 Jun 17:49] LOST CHILD2 0xfffff80204e67420->0xfffff801fe71d5a0 (actual parent 0xfffff80200c716e0) [29 Jun 17:49] Warning: The system would like to page to swap but no swap space is configured! [29 Jun 17:49] is this message very bad in the dmesg? =_) [29 Jun 17:49] (I've had to remove swap or it would crash when using swap, without swap it is months that it runs rocksolid without hangs :) [29 Jun 17:51] you should be able to ignore that warning. [29 Jun 17:51] its debugging for a code path that I need to look at [29 Jun 17:53] ok [29 Jun 17:53] good :) [29 Jun 17:54] (anyway it is not a problem even if it crashes, I have a backup of this data, but I like to push the limits ;) [29 Jun 17:54] it is a 10Tb disk with only 600Gbyte free, I will see if it will make it =_) [29 Jun 17:55] (and meanwhile bulkfree runs, transmission-daemon is downloading 500+ Gbytes ;) [29 Jun 17:57] you can kill the bulkfree if it is interfering. you can also just slow it down with a sysctl [29 Jun 17:57] sysctl vfs.hammer2.bulkfree_tps=N to slow it down, set N to like 100 or something [29 Jun 18:25] hihi [29 Jun 18:26] --> jh32 (~jh32@h1470203.stratoserver.net) joined the channel [29 Jun 18:28] yo [29 Jun 18:33] hoi [29 Jun 18:37] morning [29 Jun 19:25] dillon: ah, this slowing down is interesting, I didn't know, tnx :) [29 Jun 19:27] it finished, all seems fine: https://pastebin.com/3VWv6hcd [29 Jun 19:27] now I will run another because before running this bulkfree I did delete a 600Gbyte folder but I don't see the space reclaimed... [29 Jun 19:31] it will run those bulkfree's overnight from cron automatically [29 Jun 19:31] so unless the drive is full you can just let the overnight deal with it [29 Jun 19:40] eh overnight will be late I tried to run now to avoid the disk filling, but I will cancel more data if it doesn't yet... =_) [29 Jun 19:46] [digest-dfbsd] - pf table size check and change - https://www.dragonflydigest.com/2020/06/29/24698.html - Justin Sherrill [29 Jun 20:12] --> luxh (~luxh@Tricholoma.Update.UU.SE) joined the channel [29 Jun 21:37] the second bulkfree actually regained the space :) https://pastebin.com/wWqaHSvV [29 Jun 21:37] now I go to sleep, bbl :) [29 Jun 21:46] yo [29 Jun 21:52] --> br0nko (~br0nko@lfbn-idf1-1-1990-128.w90-127.abo.wanadoo.fr) joined the channel [29 Jun 22:20] --> Bridgeman (~damien@lfbn-mon-1-1308-3.w90-113.abo.wanadoo.fr) joined the channel [29 Jun 23:35] <-- Bridgeman (~damien@lfbn-mon-1-1308-3.w90-113.abo.wanadoo.fr) left the channel [30 Jun 01:24] hmm [30 Jun 01:24] dillon: we're going to have the binary packages a bit later this month, probably the end of this week [30 Jun 01:43] ok [30 Jun 02:34] --> aerona (~aerona@071-092-107-228.res.spectrum.com) joined the channel [30 Jun 02:41] btw 5 days uptime [30 Jun 02:41] so your patch isn't making things worse at least [30 Jun 02:47] yo, morning [30 Jun 02:51] dillon: about the make_only_dev() in tun/tap, I have two ideas: (1) make devfs_device_find_by_name() able to find devices created by make_only_dev(); (2) add cdev_t's created in {tun,tap}clone() to a list and find it from the list in {tun,tap}_clone_create() [30 Jun 02:52] dillon: what do you think about the two ideas? [30 Jun 03:28] *** You disconnected [30 Jun 04:26] moin [30 Jun 04:33] --> thomasm_ (~sinnlos@x2f45de6.dyn.telefonica.de) joined the channel [30 Jun 05:26] --> aerona (~aerona@071-092-107-228.res.spectrum.com) joined the channel [30 Jun 05:27] --> aerona (~aerona@071-092-107-228.res.spectrum.com) joined the channel [30 Jun 05:40] aly: I think just passing the dev directory in as you were doing, except make it an additional argument instead of reusing the existing argument [30 Jun 05:40] directly I mean [30 Jun 05:42] dillon: but that needs to update all callers of if_clone_create() ... [30 Jun 05:42] yes, but its the most direcf, safest solution [30 Jun 05:42] direct [30 Jun 05:42] adding another namespace or an internal special search is just too indirect [30 Jun 05:42] a direct passing of the dev is more understandable [30 Jun 05:43] yes, just passing the cdev_t is more clear [30 Jun 05:44] dillon: how about putting the cdev_t's in a local list in tun/tap ? [30 Jun 05:44] what do you mean by 'local list' ? [30 Jun 05:45] dillon: define a static SLIST in tun/tap, add cdev_t's created in {tun,tap}clone() to the list, and then find the cdev_t from the list in {tun,tap}_clone_create() [30 Jun 05:47] oh. mmm... no, I think it is unnecessary [30 Jun 05:47] I'd rather extend the API [30 Jun 05:48] dillon: ok [30 Jun 05:49] dillon: I don't know the details about freebsd's make_dev() and friends. they use 'make_dev_credf()' in {tap,tun}create() whereas we use 'make_only_dev()' [30 Jun 05:49] https://github.com/freebsd/freebsd/blame/16793845b520ce91cd81ab1ab290214f7afbd9f4/sys/net/if_tun.c#L260 [30 Jun 05:50] dillon: I'm wondering whether it is necessary to distinguish 'make_only_dev()' and 'make_dev()' ... [30 Jun 05:53] the original idea for those internal devices was that they were only used within the kernel. exposing them to userland has to be done in a very controlled manner [30 Jun 05:53] for example, the tap/tun device cannot be exposed to userland until it is actually fully initialized [30 Jun 05:53] doing it too soon and userland might be able to squeeze a command in and panic the kernel [30 Jun 05:56] dillon: thanks. I'll look at extending the if_clone_create() [30 Jun 05:57] ok. should be simple. [30 Jun 05:57] sure, thanks [30 Jun 05:59] sephe_: ^^ how about adding another argument to if_clone_create() ? [30 Jun 06:27] *** You connected [30 Jun 06:48] moin [30 Jun 07:43] --> littledra (~nobody@37.120.141.36) joined the channel [30 Jun 07:47] --> littledr2 (~nobody@37.120.141.36) joined the channel [30 Jun 07:47] *** You disconnected [30 Jun 07:47] *** You connected [30 Jun 07:47] aly: sure you can do that. in case you dont want to change every call of if_clonet_craete(), you can make a if_clone_create2(); only adjust necessary callers. howvere, if the parameter is required, then just change the if_clone_creayte interface, and adjust all callers [30 Jun 07:53] sephe_: thanks. I like the idea to add a new function if_clone_create2(); but the extra parameter need to be passed to ifc->ifc_create(), so that requires to update all modules using if_clone as well [30 Jun 07:53] any other suggestions? [30 Jun 07:55] it's just 3 consumers? why an extra function? [30 Jun 07:57] ok, then just change the interface and adjust all callers [30 Jun 07:58] ok, thanks. I'll work on a patch later. [30 Jun 07:59] sephe_: any ideas/suggestions on updating the netgraph from freebsd to gain the bluetooth support? [30 Jun 08:00] aly: we have a netbt stack from netbsd [30 Jun 08:00] aly: please also check netbsd to see whether netbt stack worth updating or not [30 Jun 08:01] aly: personally, i don't like netgraph, sorry [30 Jun 08:01] sephe_: yeah, i see the net/netbt code. just looking a bit at both sides (docs, not code yet) [30 Jun 08:04] sephe_: do you mind to give your insights on netgraph, netmap, DPDK, Cisco's VPP ? [30 Jun 08:05] (for work, we're using DPVS (DPDK-based LVS) from iQiyi for L4 load balancing) [30 Jun 08:06] go for the DPDK porting, if you want [30 Jun 08:06] --> xxavi (~xxavi@91.132.138.78) joined the channel [30 Jun 08:06] Hi to all. [30 Jun 08:06] it does have its place in the forwarding like application [30 Jun 08:07] I get this error: [30 Jun 08:07] root@DfBSD-LG# make search key=manager [30 Jun 08:07] The search target requires INDEX. Please run make index or make fetchindex. [30 Jun 08:07] for userland stack, it's crap [30 Jun 08:07] root@DfBSD-LG# make fetchindex [30 Jun 08:07] fetch: http://muscles.dragonflybsd.org/INDEX.bz2: Undefined error: 0 [30 Jun 08:07] Any idea ? [30 Jun 08:07] fetch: http://muscles.dragonflybsd.org/INDEX.bz2: Undefined error: 0 [30 Jun 08:07] ^Cfetch: transfer interrupted [30 Jun 08:08] sephe_: ok. that's a too big challenge for me, at least to my current knowledge :D [30 Jun 08:10] xxavi: how about 'make index' ? [30 Jun 08:14] Hi aly. [30 Jun 08:14] root@DfBSD-LG# make index [30 Jun 08:14] make: don't know how to make index. Stop [30 Jun 08:14] make: stopped in /usr/dports [30 Jun 08:14] root@DfBSD-LG# [30 Jun 08:16] xxavi: ok. that seems not ported from freebsd's ports ... [30 Jun 08:17] xxavi: what are you trying to do? 'pkg search' not enough for you? [30 Jun 08:17] aly: ok, thanks. [30 Jun 08:17] xxavi: I remember we support to generate the INDEX for repology. not sure. need to check with tuxillo [30 Jun 08:17] aly: Yes but, only return binaries packages and, not all Dports are in binary. [30 Jun 08:18] aly: Oh, ok, thanks. [30 Jun 08:18] xxavi: yeah, that's true [30 Jun 08:18] --> fridim (~fridim@lfbn-mon-1-707-194.w2-4.abo.wanadoo.fr) joined the channel [30 Jun 09:01] --> kworr (~Thunderbi@limbo.b1t.name) joined the channel [30 Jun 09:12] moin [30 Jun 09:17] Hi [30 Jun 09:29] --> peeterm (~peeterm@kronos.fi.tartu.ee) joined the channel