Reputation
214
Top tag
Next privilege 250 Rep.
View close votes
Badges
2 11
Newest
 Yearling
Impact
~16k people reached

  • 0 posts edited
  • 0 helpful flags
  • 99 votes cast
Oct
21
comment TCP packets larger than 4 KB don't get a reply from Linux
As the tcpdump output shows, the guest Linux receives the TCP packet with the payload of 4060 bytes, but never sends an ACK. If I reduce the the payload size to 4010 bytes, I get an ACK (shows up in the tcpdump output and the host also gets it). ICMP works back and forth with frames of 50000 bytes. Knowing all these, could you please be more specific what I should try first, and what settings related to jumbo frames I should change? Is there some information I failed to provide, or is there another debug tool I should run?
Oct
20
awarded  Commentator
Oct
20
comment TCP packets larger than 4 KB don't get a reply from Linux
I ran tcpdump -n -vv in the internal eth0. I have all packages the guest has sent or received (see them in the question). I was running User-mode Linux with eth0=slirp,,/my/own/slirp, replacing the slirp binary with my own emulator. My most important question: why isn't the guest Linux sending an ACK when the TCP payload is that large? Why does the guest Linux keep silent?
Oct
20
comment TCP packets larger than 4 KB don't get a reply from Linux
No, it shows all traffic on the guest eth0. I used the command tcpdump -n -vv without any filter.
Oct
20
comment TCP packets larger than 4 KB don't get a reply from Linux
@DavidSchwartz: There is no congestion window, my emulator sends some payload, and waits for a full ACK indefinitely. I know that a correct TCP implementation would resend eventually. But in this question I'm asking why I'm not getting an ACK from the Linux guest. Because it's not required to ACK everything is true but it doesn't answer my question: I'm interested in why it isn't ACKing in this particular case.
Oct
20
asked TCP packets larger than 4 KB don't get a reply from Linux
Oct
19
accepted pretty-printing IP packets
Oct
18
asked pretty-printing IP packets
Oct
14
revised KVM guest halt to make the host kvm exit
added 120 characters in body
Oct
14
revised KVM guest halt to make the host kvm exit
added 1694 characters in body; edited title
Oct
14
answered KVM console with scrollback
Oct
14
asked KVM console with scrollback
Oct
14
revised KVM guest halt to make the host kvm exit
added 154 characters in body
Oct
14
asked KVM guest halt to make the host kvm exit
Jan
11
awarded  Popular Question
Jul
6
awarded  Popular Question
Nov
24
comment ext4 filesystem corruption — maybe hardware error?
@HubertKario: Done.
Nov
24
revised ext4 filesystem corruption — maybe hardware error?
added 4867 characters in body
Nov
24
comment ext4 filesystem corruption — maybe hardware error?
fsck -f, when booted from the Ubuntu live installer pen drive, has found tons of errors, and fixed them. A subsequent run of fsck didn't find errors. Then, when the normal system was booted, it started emitting ext4 errors again after half an hour. No, the syslog isn't indicating any hardware errors. You can see the full dmesg output (except for the few seconds after boot) in the original question.
Nov
24
asked ext4 filesystem corruption — maybe hardware error?