Incrementing the reference count of a packet as a means of granting ownership
is unsafe when the packet is shared across gorountines. The underlying buffer's
reference count is unchanged since it "technically" has the same owning
PacketBuffer, which means different goroutines operating on the underlying
buffer (and packet itself) race.
Clones are roughly as fast as IncRefs because the PacketBuffers allocate from
a pool and the underlying buffers are cloned with copy-on-write
semantics.
I've left IncRef in places where the original packet in obviously going out of
scope at the end of the function or in some tests.
Reported-by: syzbot+e026046f4bf8ad09ae1f@syzkaller.appspotmail.com
Reported-by: syzbot+559365d6050db4b30e0f@syzkaller.appspotmail.com
Reported-by: syzbot+63c78a2c88a5744c636b@syzkaller.appspotmail.com
PiperOrigin-RevId: 705676806
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
The underlying chunks returned from PullUp should not be shared,
since the underlying slice can sometimes be directly modified. This change
also reworks some of the network parsing code so that ownership of
views is more explicit to the reader.
PiperOrigin-RevId: 538230394
The benefits of PacketBufferPtr never materialized and it makes the type
difficult to work with, e.g. it can't be used with go_generics to make a
(performant) list.
This is effectively a rollback of cl/480518221.
PiperOrigin-RevId: 530954630
A rare bug occurs when assigning a value to the reassemblers map causes a panic
due to the map being nil. The only way this is possible is if
Fragmentation.Release() is called before Process() is able to finish.
PiperOrigin-RevId: 466771159
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
Before this change the only reference on the packet after reassembly
processing was held by the reassembler in the holes array. This meant that
after the reassembly cleanup job, there were no references left on the
packet, leading to use after free bugs.
PiperOrigin-RevId: 424479461
The extra pointer indirection is not necessary and allows for a nil buffer.
This change bumps the PacketBuffer struct size from 296 to 792 bytes.
PiperOrigin-RevId: 422669812
Introduce tcpip.MonotonicTime; replace int64 in tcpip.Clock method
returns with time.Time and MonotonicTime to improve type safety and
ensure that monotonic clock readings are never compared to wall clock
readings.
PiperOrigin-RevId: 375775907
