No buffer space available on linux

What does “ping: sendmsg: No buffer space available” mean?

I had an intermittent issue (which resolves itself spontaneously, after using a different connection for a while) that caused Internet requests to fail, with one particular wireless network. After about 5 requests after associating with an AP, ping would report the error

ping: sendmsg: No buffer space available 

just as in Wifi connected but no data transfer: «ping: sendmsg: No buffer space available». In case it’s relevant, I’m seeing with a Centrino Ultimate-N 6300 [8086:4238] wireless interface on a Thinkpad X201. Bug #836250 is possibly related. While I can work around this issue, I was wondering: What does this error message mean? In particular, what buffer is it talking about?

By the hideous laughter of Kek, it turns out that having a mistyped FQDN in /etc/hostname in an Azure VM causes the network interface to behave as if it was a bad Wifi link. I have been in the «industry» for some time but it’s getting more interesting every day.

In 2023 having a modern router and ridiculous «212992» value in /proc/sys/net/core/wmem_max easily triggers this error even in a healthy laptop just during the usual apt update && apt upgrade . And all these answers with ‘disabling 5G’ or ‘putting you router in 802.11 b/g only mode’. Seriously? Go back to 2005 just because some ubuntu developers are there?

6 Answers 6

It means you reached a maximum value for a system parameter. Probably /proc/sys/net/core/wmem_max (but this might need some investigating on a system that shows this error). This setting is the maximum amount «receive socket memory».

It is likely that the cause is a broken NIC -or- a NIC that is not 100% supported if the system is not stressed out. Broadcom bcm4313 is one that seems to show this error.

In case anyone wants to know: the file where these are stored is /etc/sysctl.conf and you can alter them from a root (!) prompt. Some examples can be found here (also includes the parameter shown at the beginning of this answer). But this should only be done if replacing the NIC itself does not solve it.

I got the same issue with my laptop DELL XPS using Linux 5.3.0-42 Ubuntu 18.04.3 LTS. Using ping 8.8.8.8 I get the following error:

ping: sendmsg: No buffer space available 

Several sources online suggest to increase the wmem_max using:

echo 83886080 | sudo tee /proc/sys/net/core/wmem_max 

But I didn’t solve with this, after a while I just changed the Wireless mode in my router into

Moreover, I disabled the 5 GHz in the router, probably it is overcharged.

All of this fixed my issue.

EDIT: alternatively, if you do not want to disable 5G you can set from the wifi configuration the related wifi channel between 5G or 2.4G, just go to Wifi network settings, select your network and in the «identity» tab change the BSSID, there you should find 2 BSSID one is the 2.4GHZ and the other one is 5G.

I’m using the XPS, linux 5.4.0-60-generic, the 5.0 GHz option is disabled on my router and my wlan card is working at 2.5 GHz. I get this problem when using bluetooth earphones and wifi concurrently. Go figure.

Chances are it’s a problem with the driver or the hardware itself. It can be temporarily fixed by taking the interface up/down. Basically the output queue is full and unless you are handling some major network traffic or have a ton of hosts on the network you shouldn’t see that error, and if you were, you could probably tweak some kernel values to prevent the error.

I had a similar problem with a 3g modem (USB modem). When using a different modem, same brand, the problem disappeared. Whenever I used the defective modem it worked for a while then started to fail with same symptoms, after few days the modem broke completely, I finally replace it, and the problem solved completely.

I found that this message could be caused by a faulty network equipment somewhere on the local network, not necessarily on your computer.

For example, I got this message on my Ubuntu 20.04 when pinging local router at 192.168.1.1. In the mean time the internet traffic was interrupted when the message was present.

$ ping 192.168.1.1 . 64 bytes from 192.168.1.1: icmp_seq=216 ttl=64 time=0.295 ms 64 bytes from 192.168.1.1: icmp_seq=217 ttl=64 time=0.350 ms ping: sendmsg: No buffer space available ping: sendmsg: No buffer space available . $ ping 192.168.1.1 PING 192.168.1.1 (192.168.1.1) 56(84) bytes of data. From 192.168.1.8 icmp_seq=10 Destination Host Unreachable . 

The problem was resolved immediately by unplugging ATZEBE USB3 expansion hub used for another laptop computer from the Ethernet port on the same network switch (TP-Link) where my Ubuntu machine was plugged in. The laptop was off in the mean time. I was able to replicate the problem a few times. The problem reappeared a minute or so after I plugget the ATZEBE in again. The router has many other devices in it, but only removing the ATZEBE was the one that resolved the problem. Otherwise, the Ethernet switch might also be a suspect.

As pointed out earlier, the message is about a system parameter. You could read with:

$ cat /proc/sys/net/core/wmem_max 212992 

As the value is quite large for a simple ping, I suspect that either the faulty hardware is spamming the network and overflowing the buffer, or the system is handling a problem situation with a misleading message.

Источник

Ошибка «sendmsg: Недостаточно буферного пространства» (РЕШЕНО)

Ошибка «sendmsg: Недостаточно буферного пространства» (или если у вас в качестве языка выбран английский, то ошибка будет «ping: sendmsg: No buffer space available») может возникнуть при использовании сети в Linux.

Проблема может быть периодической, то есть Интернет-подключение может работать нормально, но иногда «зависать». Чаще возникает на Wi-Fi подключениях, а не на беспроводных сетевых интерфейсах. Обычно помогает включение и отключение Wi-Fi адаптера. К сожалению, не всегда эту проблему можно решить изменив настройки системы — в некоторых случаях это признак аппаратной проблемы сетевого адаптера или неполной совместимости.

Пример команды, которая может вызвать:

PING 192.168.0.1 (192.168.0.1) 56(84) bytes of data. ping: sendmsg: Недостаточно буферного пространства ping: sendmsg: Недостаточно буферного пространства ping: sendmsg: Недостаточно буферного пространства ping: sendmsg: Недостаточно буферного пространства

Данная ошибка означает, что вы достигли максимального значения системного параметра. Вероятно, /proc/sys/net/core/wmem_max (но это может потребовать некоторого исследования в системе, которая показывает эту ошибку). Этот параметр представляет собой максимальный объем «принимаемой памяти сокета».

Вполне вероятно, что причиной является неисправный сетевой адаптер или сетевой адаптер, который не поддерживается на 100%, но в принципе работает, если система не перегружена.

На случай, если кто-то захочет узнать: файл, в котором хранятся указанные настройки, — это /etc/sysctl.conf, и вы можете изменить их — для этого нужны права root. Некоторые примеры можно найти здесь (также включает параметр, указанный в начале этого ответа). Но это следует делать только в том случае, если замена самой сетевой карты не решает эту проблему.

Скорее всего, проблема в драйвере или в самом оборудовании. Это можно временно исправить, отключив/включив интерфейс. В основном очередь вывода заполнена, и если вы не обрабатываете какой-либо основной сетевой трафик или не имеете тонны хостов в сети, вы не должны видеть эту ошибку.

Некоторым пользователям удаётся решить эту проблему изменив настройки Wi-Fi в роутере, например, выбрав режим «802.11 b/g/n mixed».

Иногда проблему удаётся решить только заменив дефектный модем или беспроводной адаптер.

В качестве быстрого исправления вы можете выключить/включить Wi-Fi адаптер в NetwordManager или с помощью команд:

sudo ip link set ИНТЕРФЕЙС down sudo ip link set ИНТЕРФЕЙС up

Если вы хотите разобраться в причинах и корне данной проблемы, то обратитесь к статье «Памятка по Wi-Fi в Linux: подсказки и решение проблем», найдите там раздел «Как увидеть, что происходит с беспроводными интерфейсами» и используйте указанные команды.

Связанные статьи:

Источник

PEAK-System Forum

Socket write avoiding No buffer space available

Socket write avoiding No buffer space available

Post by agecheris » Fri 14. Apr 2017, 14:26

Hi,
I am using the PCAN adapter with the 8.1 Linux driver in Ubuntu 16.04. I have implemented SocketCAN support in my application but when writing I get errno=105 «No buffer space available». How can I avoid this?

I have been reading about different calls of setsockopt, ioctl etc but I cannot find any information on how to determine the free space of the write buffer. Any help is appreciated.

Re: Socket write avoiding No buffer space available

Post by S.Grosjean » Wed 19. Apr 2017, 10:03

There’s many reasons why the socketCAN layer cannot flush its internal buffers.
First of all, please give us the result of «cat /proc/pcan» as well as «uname -a».

You might also install the last version of pcan from here:
http://www.peak-system.com/fileadmin/me . 3.1.tar.gz

Re: Socket write avoiding No buffer space available

Post by agecheris » Wed 19. Apr 2017, 18:31

Hi Stéphane,
Thank you for the reply. Here comes the requested information.
I dont think there is any problem with the Peak related stuff.
I just want to know how the determine the state of the tx buffer.
This is maybe more related to SocketCAN than Peak.

If I increase the tx buffer like this «ifconfig can0 up txqueuelen 500»
I dont get the 105 error. It seems to default to 10. I guess that it is 10 frames.
But in my application I want to be able to handle the case when the tx buffer gets full whatever the its size is.

cat /proc/pcan
*————- PEAK-System CAN interfaces (www.peak-system.com) ————-
*————- Release_20170214_n (8.3.1) Apr 19 2017 20:28:49 —————
*———- [mod] [isa] [pci] [pec] [dng] [par] [usb] [pcc] [net] ————
*——————— 1 interfaces @ major 246 found ————————
*n -type- -ndev- —base— irq —btr- —read— —write- —irqs— -errors- status
32 usb can0 ffffffff 002 0x001c 00000000 00000000 00000000 00000000 0x0000

uname -a
Linux mikael-VirtualBox 4.4.0-72-generic #93-Ubuntu SMP Fri Mar 31 14:07:23 UTC 2017 i686 i686 i686 GNU/Linux

Re: Socket write avoiding No buffer space available

Post by agecheris » Wed 19. Apr 2017, 20:40

With the latest driver I dont get the error, even with txqueuelen=1. But my request remain. I would like to write robust code but I dont know how to get the status of the tx buffer.

Re: Socket write avoiding No buffer space available

Post by M.Maidhof » Thu 20. Apr 2017, 09:53

do you use a virtual machine? Then this buffer problem could be related to your Virtual Box USB system Please check that all “guest extension packs” are installed on that VM. We don´t support VM related problems. So please test again in a native Linux system, to see if that problem is based on the virtual system.

Re: Socket write avoiding No buffer space available

Post by S.Grosjean » Thu 20. Apr 2017, 10:11

Thx for your feedback. AFAIU you’re running a Linux system into a virtual box, with a PCAN-USB device connected to. We generally do not support such configuration and generally strongly advice to move on a real Linux machine.

To answer to your request: socket-CAN is (obviously) a socket-based API. There is actually no way to determine the status of the Tx buffer, because socket are historically IP medium oriented. Using the sockets with IP protocols means: 1/ using UDP and thus, knowing that some datagrams may be lost 2/ using TCP and thus, trusting in the (socket) protocol layers to handle lost and/or out-of-order packets. In both cases, Tx buffer size is no longer an issue.

Thus, the only way of controlling CAN frames sending is to handle errno=105 by yourself, then, setting up some (not busy) wait cycles before retrying the same «send()» command. You can have a look to the cansend.c application from the can-utils package to learn more about that.

Источник