Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

net-test: packetdrill: update tcp/epoll tests to latest kernel behavior #76

Open
wants to merge 1 commit into
base: master
Choose a base branch
from
Open

net-test: packetdrill: update tcp/epoll tests to latest kernel behavior #76

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
25 changes: 15 additions & 10 deletions gtests/net/tcp/epoll/epoll_out_edge.pkt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -34,21 +34,24 @@
// Now epoll_wait should not return any event because there is no space.
+0 epoll_wait(5, {events=0, ptr=0}, 1, 0) = 0

// The ack does not open up enough space to send EPOLLOUT to the user.
// The ack opens up some space, and that will send EPOLLOUT to the user.
// Before kernel commit 4bfe744ff164 "tcp: fix potential xmit stalls caused
// by TCP_NOTSENT_LOWAT", it would not send EPOLLOUT, and that was a bug.
+0 > P. 1:10001(10000) ack 1
+0 < . 1:1(0) ack 10001 win 257
+0 > P. 10001:26001(16000) ack 1
+0 epoll_wait(5, {events=0, ptr=0}, 1, 0) = 0
+0 epoll_wait(5, {events=EPOLLOUT, fd=4}, 1, 0) = 1

// Now this ack opens up substantial space and we will send EPOLLOUT.
// Ack what we received so far. This frees up additional queue space but
// does not change the state because the queue was not full before.
+0 < . 1:1(0) ack 26001 win 257
+0 > P. 26001:42001(16000) ack 1
+0 epoll_wait(5, {events=EPOLLOUT, fd=4}, 1, 0) = 1
// EPOLLET prevents a second EPOLLOUT.
+0 epoll_wait(5, {events=0, ptr=0}, 1, 0) = 0

// We have another parital write here.
+0 write(4, ..., 180000) = 128000
// We have another parital write here. This fills up the queue and so the
// state will change after we send the next ack and free up some space.
+0 write(4, ..., 180000) = 64000

// The ack opens enough space to send EPOLLOUT to the user.
+0 < . 1:1(0) ack 42001 win 257
Expand All @@ -61,10 +64,12 @@
+0 epoll_wait(5, {events=0, ptr=0}, 1, 0) = 0
+0 > P. 74001:106001(32000) ack 1

// Another partial write.
+0 write(4, ..., 180000) = 112000
// Do a full write this time. This will not fill up the queue.
+0 write(4, ..., 180000) = 180000

// This ack opens more space, so we send EPOLLOUT.
// This ack opens more space, but EPOLLET prevents a second EPOLLOUT
// because the previous write didn't fill up the queue and so the state
// hasn't changed.
+0 < . 1:1(0) ack 106001 win 257
+0 > P. 106001:138001(32000) ack 1
+0 epoll_wait(5, {events=EPOLLOUT, fd=4}, 1, 0) = 1
+0 epoll_wait(5, {events=0, ptr=0}, 1, 0) = 0
25 changes: 15 additions & 10 deletions gtests/net/tcp/epoll/epoll_out_edge_default_notsent_lowat.pkt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -35,21 +35,24 @@
// Now epoll_wait should not return any event because there is no space.
+0 epoll_wait(5, {events=0, ptr=0}, 1, 0) = 0

// The ack does not open up enough space to send EPOLLOUT to the user.
// The ack opens up some space, and that will send EPOLLOUT to the user.
// Before kernel commit 4bfe744ff164 "tcp: fix potential xmit stalls caused
// by TCP_NOTSENT_LOWAT", it would not send EPOLLOUT, and that was a bug.
+0 > P. 1:10001(10000) ack 1
+0 < . 1:1(0) ack 10001 win 257
+0 > P. 10001:26001(16000) ack 1
+0 epoll_wait(5, {events=0, ptr=0}, 1, 0) = 0
+0 epoll_wait(5, {events=EPOLLOUT, fd=4}, 1, 0) = 1

// Now this ack opens up substantial space and we will send EPOLLOUT.
// Ack what we received so far. This frees up additional queue space but
// does not change the state because the queue was not full before.
+0 < . 1:1(0) ack 26001 win 257
+0 > P. 26001:42001(16000) ack 1
+0 epoll_wait(5, {events=EPOLLOUT, fd=4}, 1, 0) = 1
// EPOLLET prevents a second EPOLLOUT.
+0 epoll_wait(5, {events=0, ptr=0}, 1, 0) = 0

// We have another parital write here.
+0 write(4, ..., 180000) = 128000
// We have another parital write here. This fills up the queue and so the
// state will change after we send the next ack and free up some space.
+0 write(4, ..., 180000) = 64000

// The ack opens enough space to send EPOLLOUT to the user.
+0 < . 1:1(0) ack 42001 win 257
Expand All @@ -62,10 +65,12 @@
+0 epoll_wait(5, {events=0, ptr=0}, 1, 0) = 0
+0 > P. 74001:106001(32000) ack 1

// Another partial write.
+0 write(4, ..., 180000) = 112000
// Do a full write this time. This will not fill up the queue.
+0 write(4, ..., 180000) = 180000

// This ack opens more space, so we send EPOLLOUT.
// This ack opens more space, but EPOLLET prevents a second EPOLLOUT
// because the previous write didn't fill up the queue and so the state
// hasn't changed.
+0 < . 1:1(0) ack 106001 win 257
+0 > P. 106001:138001(32000) ack 1
+0 epoll_wait(5, {events=EPOLLOUT, fd=4}, 1, 0) = 1
+0 epoll_wait(5, {events=0, ptr=0}, 1, 0) = 0
10 changes: 6 additions & 4 deletions gtests/net/tcp/epoll/epoll_out_edge_notsent_lowat.pkt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -34,16 +34,18 @@
// Now epoll_wait should not return any event because there is no space.
+0 epoll_wait(5, {events=0, ptr=0}, 1, 0) = 0

// The ack does not open up enough space to send EPOLLOUT to the user.
// The ack opens up some space, and that will send EPOLLOUT to the user.
// Before kernel commit 4bfe744ff164 "tcp: fix potential xmit stalls caused
// by TCP_NOTSENT_LOWAT", it would not send EPOLLOUT, and that was a bug.
+0 > P. 1:10001(10000) ack 1
+0 < . 1:1(0) ack 10001 win 257
+0 > P. 10001:26001(16000) ack 1
+0 epoll_wait(5, {events=0, ptr=0}, 1, 0) = 0
+0 epoll_wait(5, {events=EPOLLOUT, fd=4}, 1, 0) = 1

// Now this ack opens up substantial space and we will send EPOLLOUT.
// Ack what we received so far. This frees up additional queue space but
// does not change the state because the queue was not full before.
+0 < . 1:1(0) ack 26001 win 257
+0 > P. 26001:42001(16000) ack 1
+0 epoll_wait(5, {events=EPOLLOUT, fd=4}, 1, 0) = 1
// EPOLLET prevents a second EPOLLOUT.
+0 epoll_wait(5, {events=0, ptr=0}, 1, 0) = 0

Expand Down