`//pkg/tcpip/link/rawfile` is a package to deal with raw socket and file FDs.
It is not only used for Netstack, but rather just generally useful raw file
manipulation stuff.
This change removes the Unix-error-to-`//pkg/tcpip`-error translation step
from its functions; this is now the responsibility of its callers. Callers
within Netstack now do the translation by themselves; the translation
function is moved to `//pkg/tcpip`.
This allows the `//pkg/tcpip/link/rawfile` package to not depend on
`//pkg/tcpip`, which in turn means the `//pkg/eventfd` package
(which depends on `rawfile`) no longer transitively depends on
`//pkg/tcpip`, which in turns means the `//pkg/unet` package (which
depends on `//pkg/eventfd`) no longer transitively depends on
`//pkg/tcpip`, which in turns means that the `//pkg/eventchannel`
package (which depends on `//pkg/unet`) no longer transitively
depends on `//pkg/tcpip`, which in turns means that the `//pkg/metric`
package (which depends on `//pkg/eventchannel`) no longer transitively
depends on `//pkg/tcpip`, which finally means that the `//pkg/metric`
package can be used within `//pkg/tcpip`. \o/
This changes does not make it use it, it just moves `rawfile`.
PiperOrigin-RevId: 647943618
Marks the structs in netstack as savable. This does not change or break any
existing behavior as the netstack itself is not savable yet.
PiperOrigin-RevId: 635943481
It is an idea of running codespell as part of our presubmit checks.
Before enabling it for new changes, let's fix what it has found.
Signed-off-by: Andrei Vagin <avagin@gmail.com>
Previously, when a packet was sent from a raw packet socket, the packet
buffer's link header was left unpopulated and the link header was only
found in the packet buffer's payload. This breaks the expectations of
LinkEndpoints which expect the link layer header to always be populated
when the link requires a header.
PiperOrigin-RevId: 542349445
Really, any LinkWriter is compatible with gVisor GSO: there's nothing special
to be done. Anything with a WritePackets method is good to go.
PiperOrigin-RevId: 537048401
This is effectively a rollback of cl/450976957. The original motivation never
panned out, and it's easier to work with the lists. They also are easier to
avoid allocations with.
PiperOrigin-RevId: 531020857
The dispatch loop could pickup the change to e.stopRequested and close the
eventFD before the call to Notify().
See bug for more detail.
PiperOrigin-RevId: 507626752
This change has significant performance implications. bufferv2 is reference
counted and pooled, which alleviates heap/GC pressure. Below are the results
from running the iperf benchmark.
HEAD:
BenchmarkIperf/operation.Upload-16 1552 ns/op 46.6GiB total allocations
BenchmarkIperf/operation.Download-16 1114 ns/op 68.6GiB total allocations
w/ change:
BenchmarkIperf/operation.Upload-16 1139 ns/op (-27%) 1.41GiB total allocations (-97%)
BenchmarkIperf/operation.Download-16 753.2 ns/op (-33%) 706MiB total allocations (-99%)
PiperOrigin-RevId: 462453185
Andrei VaginandgVisor bot
Etienne PerotandgVisor bot
Jing ChenandgVisor bot
Nayana BidariandgVisor bot
Kevin KrakauerandgVisor bot
Andrei Vagin
Ghanan GowripalanandgVisor bot
Lucas ManningandgVisor bot