The changes below src/crypto/internal/fips/ are mechanical.
See fipsdeps.go and fipsdeps_test.go for the rationale.
For #69536
Change-Id: I292ce65237cd8d2fd87fab99814514dd0e69c4a7
Reviewed-on: https://go-review.googlesource.com/c/go/+/627956
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
Reviewed-by: Daniel McCarney <daniel@binaryparadox.net>
Auto-Submit: Filippo Valsorda <filippo@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
As explained in fips_test.go, we generally want to minimize tests inside
the FIPS module. When there is a relevant calling package, the tests
should go there, otherwise in fipstest.
This required redoing a bit the CAST failure tests, but the new version
is actually more robust because it will fail if a _ import is missing.
Since TestCAST doesn't print a line for each passed CAST anymore, made
GODEBUG=fips140=debug do that, in case we need to show it to the lab.
For #69536
Change-Id: I0c1b82a4e9ee39e8df9bbe95bebb0527753f51c8
Reviewed-on: https://go-review.googlesource.com/c/go/+/627955
Reviewed-by: Daniel McCarney <daniel@binaryparadox.net>
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
Auto-Submit: Filippo Valsorda <filippo@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Add new func Unzip, which is Download (= download+unzip)
without the download. This will be used for unpacking the
FIPS module zips, which are part of the Go distribution,
not downloaded.
Change-Id: Ia04f8c376db8cb1cf27acb5567dd256afb10f410
Reviewed-on: https://go-review.googlesource.com/c/go/+/629200
Auto-Submit: Russ Cox <rsc@golang.org>
Reviewed-by: Michael Matloob <matloob@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
GOFIPS140 does two things: (1) control whether to build binaries that
run in FIPS-140 mode by default, and (2) control which version of the
crypto/internal/fips source tree to use during a build.
This CL implements part (1). It recognizes the GOFIPS140 settings
"off" and "latest" and uses them to set the default GODEBUG=fips140
setting to "off" or "on" accordingly.
The documentation for GOFIPS140 is in a follow-up CL.
See cmd/go/internal/fips/fips.go for an overview.
For #70200.
Change-Id: I045f8ae0f19778a1e72a5cd2b6a7b0c88934fc30
Reviewed-on: https://go-review.googlesource.com/c/go/+/629198
Auto-Submit: Russ Cox <rsc@golang.org>
Reviewed-by: Michael Matloob <matloob@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Hashing the key means we have to take the address of it. That inhibits
subsequent optimizations on the key variable.
By hashing a copy, we incur an extra store at the hash callsite, but
we no longer need a load of the key in the inner loop. It can live
in a register throughout. (Technically, it gets spilled around
the call to the hasher, but it gets restored outside the loop.)
Maybe one day we can have special hash functions that take
int64/int32/string instead of *int64/*int32/*string.
Change-Id: Iba3133f6e82328f53c0abcb5eec13ee47c4969d1
Reviewed-on: https://go-review.googlesource.com/c/go/+/629419
Reviewed-by: Russ Cox <rsc@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Pratt <mpratt@google.com>
Note this doesn't work with int32 keys because alignment padding can change
the offset of the element.
Change-Id: I27804d3cfc7cc1b7f995f7e29630f0824f0ee899
Reviewed-on: https://go-review.googlesource.com/c/go/+/629418
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Pratt <mpratt@google.com>
Reviewed-by: Russ Cox <rsc@golang.org>
This reduces the adds required at the return point from 3 to 1.
(The multiply inside g.elem() does get CSE'd with the one inside
g.key(), but the rest of the adds don't.)
Instead, compute the element as just a fixed offset from the key.
Change-Id: Ia4d7664efafcdca5e9daeb77d270651bb186232c
Reviewed-on: https://go-review.googlesource.com/c/go/+/629535
Reviewed-by: Russ Cox <rsc@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Pratt <mpratt@google.com>
TestAllocations is failing on PPC64 causing all PPC64 CI to fail.
Skip the test until it can be debugged.
For #70448
Change-Id: Ic18e402f4af5939a90eba2e1f2b182699013ed55
Reviewed-on: https://go-review.googlesource.com/c/go/+/629697
Reviewed-by: Filippo Valsorda <filippo@golang.org>
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
Decided to automatically duplicate the high-level code to avoid growing
the ML-KEM-768 data structures.
For #70122
Change-Id: I5c705b71ee1e23adba9113d5cf6b6e505c028967
Reviewed-on: https://go-review.googlesource.com/c/go/+/621983
Auto-Submit: Filippo Valsorda <filippo@golang.org>
Reviewed-by: Roland Shoemaker <roland@golang.org>
Reviewed-by: Daniel McCarney <daniel@binaryparadox.net>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
Instead of open-coding the waitid syscall wrapper add it to
internal/syscall/unix. As the syscall is currently only used on Linux,
switch the implementation in os.(*Process).blockUntilWaitable to use the
128-byte unix.SiginfoChild type instead of a plain 128-byte buffer.
Also use ignoringEINTR for the waitid calls instead of open-coding it.
Change-Id: I8dc47e361faa1f5e912d5de021f119c91c9f12f5
Reviewed-on: https://go-review.googlesource.com/c/go/+/629655
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
Reviewed-by: Ian Lance Taylor <iant@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Tobias Klauser <tobias.klauser@gmail.com>
Check that all the FIPS zips build.
Change-Id: Iec22d9295178f95862060e57a8ac9ed657f69943
Reviewed-on: https://go-review.googlesource.com/c/go/+/629197
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Matloob <matloob@golang.org>
It never worked before, an apparent oversight.
This will also make tests work, since tests run vet.
The new FIPS mode will use overlays, so this was
keeping go test from working in that mode.
Fixes#44957.
Change-Id: I3a77846b3b3f9a3a53118aaece93ee93214a36a6
Reviewed-on: https://go-review.googlesource.com/c/go/+/629199
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Matloob <matloob@golang.org>
Add an internal (for now) implementation of testing/synctest.
The synctest.Run function executes a tree of goroutines in an
isolated environment using a fake clock. The synctest.Wait function
allows a test to wait for all other goroutines within the test
to reach a blocking point.
For #67434
For #69687
Change-Id: Icb39e54c54cece96517e58ef9cfb18bf68506cfc
Reviewed-on: https://go-review.googlesource.com/c/go/+/591997
Reviewed-by: Michael Pratt <mpratt@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Avoids a nosplit stack overflow on OpenBSD after CL 591997
increases the adjustSignalStack stack by 16 bytes.
Change-Id: I2c990de6c7cd8d2aca6e6b98133da120c8a4174b
Reviewed-on: https://go-review.googlesource.com/c/go/+/629696
Reviewed-by: Cherry Mui <cherryyz@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
In the process, replace out-of-module imports with their FIPS versions.
For #69536
Change-Id: I83e900b7c38ecf760382e5dca7fd0b1eaa5a5589
Reviewed-on: https://go-review.googlesource.com/c/go/+/626879
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Russ Cox <rsc@golang.org>
Auto-Submit: Filippo Valsorda <filippo@golang.org>
Reviewed-by: Daniel McCarney <daniel@binaryparadox.net>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
In preparation for introducing ML-KEM-1024.
Aside from the constants at the top, all other changes were automated.
Change-Id: I0fafce9a776c7b0b9179be1c858709cabf60e80f
Reviewed-on: https://go-review.googlesource.com/c/go/+/621981
Auto-Submit: Filippo Valsorda <filippo@golang.org>
Reviewed-by: Roland Shoemaker <roland@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
Reviewed-by: Daniel McCarney <daniel@binaryparadox.net>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
This will make it easier to support multiple sizes if needed.
Change-Id: I47495559fdbbf678fd98421ad6cb28172e5c810d
Reviewed-on: https://go-review.googlesource.com/c/go/+/621977
Reviewed-by: Daniel McCarney <daniel@binaryparadox.net>
Reviewed-by: Russ Cox <rsc@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Filippo Valsorda <filippo@golang.org>
Reviewed-by: Roland Shoemaker <roland@golang.org>
Main changes are the domain separator in KeyGen, dropping support for
the extended decapsulation key encoding in favor of seeds (see
https://words.filippo.io/ml-kem-seeds/), and documentation changes.
Change-Id: I8773034929094b3242aa39ac6e9d007c2bc94e63
Reviewed-on: https://go-review.googlesource.com/c/go/+/621975
Reviewed-by: Daniel McCarney <daniel@binaryparadox.net>
Auto-Submit: Filippo Valsorda <filippo@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
Reviewed-by: Roland Shoemaker <roland@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
During the development of error wrapping (#29934),
the tests were modified to stop using reflect.DeepEqual
since the prototype for error wrapping at the time included
frame information of where the error was created.
However, that change diminished the fidelity of the test
so that it is no longer as strict, which affects the endeavor
to implement v1 in terms of the v2 prototype.
For locally declared error types, use reflect.DeepEqual
to check that the exact structure of the error value matches.
Change-Id: I443d418533866ab8d533bca3785fdc741e2c140e
Reviewed-on: https://go-review.googlesource.com/c/go/+/629517
Auto-Submit: Damien Neil <dneil@google.com>
Reviewed-by: Ian Lance Taylor <iant@google.com>
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
Reviewed-by: Damien Neil <dneil@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
GOFIPS140 will be used to control whether to build binaries that
run in FIPS-140 mode by default, as well as which version of
crypto/internal/fips is used during a given build.
It is a target configuration variable analogous to
GOOS, GOARCH, CGO_ENABLED, and the like, so the
default value is recorded in the toolchain during make.bash.
This CL adds the GOFIPS140 setting to the build process
and records the default for use by cmd/go.
For #70200.
Change-Id: Iafcb5a4207f00fae8bcd93e0184a63c72526abea
Reviewed-on: https://go-review.googlesource.com/c/go/+/629196
Reviewed-by: Michael Matloob <matloob@golang.org>
Auto-Submit: Russ Cox <rsc@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
fsys.Bind(repl, dir) makes the virtual file system
redirect any references to dir to use repl instead.
In Plan 9 terms, it binds repl onto dir.
In Linux terms, it does a mount --bind of repl onto dir.
Or think of it as being like a symlink dir -> repl being
added to the virtual file system.
This is a separate layer from the overlay so that editors
working in the replacement directory can still apply
their own replacements within that tree, and also so
that editors working in the original dir do not have any
effect at all.
(If the binds and the overlay were in the same sorted list,
we'd have problems with keeping the relative priorities
of individual entries correct.)
Change-Id: Ibc88021cc95a3b8574efd5f37772ccb723aa8f7b
Reviewed-on: https://go-review.googlesource.com/c/go/+/628702
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Russ Cox <rsc@golang.org>
Reviewed-by: Michael Matloob <matloob@golang.org>
Code like x := [12]byte{1,2,3,4,5,6,7,8,9,10,11,12} stores x in
a pair of registers and uses MOVD/MOVWU to load the values
from RODATA. The code generator needs to understand not
to use the aligned PC-relative relocation for that sequence.
In non-FIPS modes, more statictemp optimizations can be applied
and this problematic sequence doesn't happen.
Fix the decision about whether to assume alignment to match
the code used by the linker when deciding what to align.
Fixes the linker failure in CL 626437 patch set 5.
Change-Id: Iedad862c6faee758d4a2c5120cab2d329265b134
Reviewed-on: https://go-review.googlesource.com/c/go/+/628835
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Mui <cherryyz@google.com>
TryBot-Bypass: Russ Cox <rsc@golang.org>
The new implementation encodes the key schedule into the type system,
which is actually nicer than what we had before.
For #69536
Change-Id: Iddab62c2aae40bc2425a155443576bb9b7aafe03
Reviewed-on: https://go-review.googlesource.com/c/go/+/626836
Reviewed-by: Russ Cox <rsc@golang.org>
Reviewed-by: Roland Shoemaker <roland@golang.org>
Commit-Queue: Roland Shoemaker <roland@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Filippo Valsorda <filippo@golang.org>
Reviewed-by: Daniel McCarney <daniel@binaryparadox.net>
This reverts commit CL 618496.
Reason for revert: https://github.com/golang/go/issues/54040#issuecomment-2485151973
Change-Id: I3bf27f7fdd475a005cb6aa190994153504e96fb5
Reviewed-on: https://go-review.googlesource.com/c/go/+/629595
Reviewed-by: Ian Lance Taylor <iant@google.com>
Auto-Submit: Ian Lance Taylor <iant@google.com>
Reviewed-by: Quim Muntal <quimmuntal@gmail.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Damien Neil <dneil@google.com>
The race detector doesn't like that windows.GetSid* functions return
pointers to the SID structure. This change makes these functions return
values instead and mark them with nocheckptr.
Fixes#70378
Cq-Include-Trybots: luci.golang.try:gotip-windows-amd64-race
Change-Id: Iba39d75bb31679d25a5ee43b51e4abb0c435dbac
Reviewed-on: https://go-review.googlesource.com/c/go/+/628995
Reviewed-by: Cherry Mui <cherryyz@google.com>
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Adds the Recipient role, alongside the existing Sender role. Also factor
out all of the shared underlying bits and pieces into a shared type that
is embedded in the Sender/Recipient roles.
Change-Id: I7640d8732aa0dd5cc9e38b8c26f0cfa7856170f6
Reviewed-on: https://go-review.googlesource.com/c/go/+/623575
Auto-Submit: Roland Shoemaker <roland@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Daniel McCarney <daniel@binaryparadox.net>
Reviewed-by: Filippo Valsorda <filippo@golang.org>
Add a new function, WithDataIndependentTiming, which takes a function as
an argument, and encloses it with calls to set/unset the DIT PSTATE bit
on Arm64.
Since DIT is OS thread-local, for the duration of the execution of
WithDataIndependentTiming, we lock the goroutine to the OS thread, using
LockOSThread. For long running operations, this is likely to not be
performant, but we expect this to be tightly scoped around cryptographic
operations that have bounded execution times.
If locking to the OS thread turns out to be too slow, another option is
to add a bit to the g state indicating if a goroutine has DIT enabled,
and then have the scheduler enable/disable DIT when scheduling a g.
Additionally, we add a new GODEBUG, dataindependenttiming, which allows
setting DIT for an entire program. Running a program with
dataindependenttiming=1 enables DIT for the program during
initialization. In an ideal world PSTATE.DIT would be inherited from
the parent thread, so we'd only need to set it in the main thread and
then all subsequent threads would inherit the value. While this does
happen in the Linux kernel [0], it is not the case for darwin [1].
Rather than add complex logic to only set it on darwin for each new
thread, we just unconditionally set it in mstart1 and cgocallbackg1
regardless of the OS. DIT will already impose some overhead, and the
cost of setting the bit is only ~two instructions (CALL, MSR), so it
should be cheap enough.
Fixes#66450
Updates #49702
[0] e8bdb3c8be/arch/arm64/kernel/process.c (L373)
[1] 8d741a5de7/osfmk/arm64/status.c (L1666)
Change-Id: I78eda691ff9254b0415f2b54770e5850a0179749
Reviewed-on: https://go-review.googlesource.com/c/go/+/598336
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Filippo Valsorda <filippo@golang.org>
Reviewed-by: Michael Pratt <mpratt@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Replace the tree of nodes with a sorted list of file replacements.
The most important property of this representation is that it
allows replacing directories: a replacement x -> y where y is
a directory could not be implemented before, because it would
require making a node for every file in the tree rooted at y,
or else it would require unsuccessful lookups for files like
x/a/b/c/d/e/f/g/h/i/j/k to try every possible parent in order
to discover the x -> y mapping.
The sorted list makes it easy to find the x -> y mapping:
when you do the binary search for x/a/b/c/d/e/f/g/h/i/j/k,
you end up immediately after the x -> y mapping, so stepping
backward one entry provides the mapping we need, if it exists.
This CL does not allow overlay files to include directories,
but now it is possible. This is at least useful for other kinds
of experiments (like FIPS).
Change-Id: Ief0afaee82e644dab8ae4eafeec20440afee2e36
Reviewed-on: https://go-review.googlesource.com/c/go/+/628701
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Matloob <matloob@golang.org>
Tests imported from x/crypto, but the actual implementation was simpler
to implement ex-novo with a #61477-like API.
Updates #61477
For #69536
Change-Id: I5a9e8a71d8abd5b2aa6b74e73bf7f631ed0115cd
Reviewed-on: https://go-review.googlesource.com/c/go/+/621275
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
Auto-Submit: Filippo Valsorda <filippo@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Daniel McCarney <daniel@binaryparadox.net>
Reviewed-by: Russ Cox <rsc@golang.org>
These are most likely redundant, but cmd/compile/internal/amd64's
TestGoAMD64v1 turns them off when clobbering those instructions, so we
need to know to skip the assembly in those cases.
Thankfully we have Avo now that adds a helpful comment with the list of
features used by each generated function!
Also improve the error output of TestGoAMD64v1. It had broken before in
#49402 and had required the exact same patch.
Change-Id: I7fab8f36042cdff630f806723aa1d8124c294f60
Reviewed-on: https://go-review.googlesource.com/c/go/+/626876
Auto-Submit: Filippo Valsorda <filippo@golang.org>
Reviewed-by: Keith Randall <khr@google.com>
Reviewed-by: Russ Cox <rsc@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Daniel McCarney <daniel@binaryparadox.net>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Do the overlay consistency checks separate from constructing
the overlay data structure. This makes sure that the data structure
can be changed without worrying about losing the checks.
Change-Id: I9ff50cc366b5362adc5570f94e6caf646ddf5046
Reviewed-on: https://go-review.googlesource.com/c/go/+/628700
Auto-Submit: Russ Cox <rsc@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Matloob <matloob@golang.org>
Refactor vfs lookup into 'func stat', which knows the internal
data structures for the vfs and returns information about a
given path. The callers can then all use stat and avoid direct
knowledge of the internal data structures.
This is setting up for a different internal data structure.
Change-Id: I496b7b3fb686cdde81b14687f65eb0bf51ec62be
Reviewed-on: https://go-review.googlesource.com/c/go/+/628699
Auto-Submit: Russ Cox <rsc@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Matloob <matloob@golang.org>
Many releases ago we migrated
from ioutil.ReadDir, which returned []os.FileInfo,
to os.ReadDir, which returns []fs.DirEntry.
The latter is faster, but the former is expected by go/build.Context.
Convert fsys to use the new ReadDir signature.
This should make the go command faster when scanning
source trees, and it brings cmd/go up to date with the rest
of the tree.
Similarly, convert Walk to WalkDir.
Change-Id: I767a8548d7ca7cc3c05f2ff073d18070a4e8a0da
Reviewed-on: https://go-review.googlesource.com/c/go/+/628698
Auto-Submit: Russ Cox <rsc@golang.org>
Reviewed-by: Michael Matloob <matloob@golang.org>
TryBot-Bypass: Russ Cox <rsc@golang.org>
Reviewed-by: Sam Thanawalla <samthanawalla@google.com>
Rename canonicalize to abs.
Rename IsDirWithGoFiles to IsGoDir.
Remove Init argument.
Split OverlayPath into Actual and Renamed.
Clean up doc comments.
Other minor cleanups.
Preparation for larger changes.
Change-Id: Ida022588149a1618a63acc91e3800b09df873b6e
Reviewed-on: https://go-review.googlesource.com/c/go/+/628697
TryBot-Bypass: Russ Cox <rsc@golang.org>
Auto-Submit: Russ Cox <rsc@golang.org>
Reviewed-by: Michael Matloob <matloob@golang.org>
