Dmitriy Vyukov [Thu, 20 Nov 2014 14:51:02 +0000 (09:51 -0500)]
runtime: fix atomic operations on non-heap addresses
Race detector runtime does not tolerate operations on addresses
that was not previously declared with __tsan_map_shadow
(namely, data, bss and heap). The corresponding address
checks for atomic operations were removed in
https://golang.org/cl/111310044
Restore these checks.
It's tricker than just not calling into race runtime,
because it is the race runtime that makes the atomic
operations themselves (if we do not call into race runtime
we skip the atomic operation itself as well). So instead we call
__tsan_go_ignore_sync_start/end around the atomic operation.
This forces race runtime to skip all other processing
except than doing the atomic operation itself.
Fixes #9136.
Russ Cox [Wed, 19 Nov 2014 20:25:33 +0000 (15:25 -0500)]
runtime: remove assumption that noptrdata data bss noptrbss are ordered and contiguous
The assumption can be violated by external linkers reordering them or
inserting non-Go sections in between them. I looked briefly at trying
to write out the _go_.o in external linking mode in a way that forced
the ordering, but no matter what there's no way to force Go's data
and Go's bss to be next to each other. If there is any data or bss from
non-Go objects, it's very likely to get stuck in between them.
Instead, rewrite the two places we know about that make the assumption.
I grepped for noptrdata to look for more and didn't find any.
The added race test (os/exec in external linking mode) fails without
the changes in the runtime. It crashes with an invalid pointer dereference.
Austin Clements [Wed, 19 Nov 2014 19:56:49 +0000 (14:56 -0500)]
[dev.cc] runtime: add explicit siginfo.si_addr field
struct siginfo_t's si_addr field is part of a union.
Previously, we represented this union in Go using an opaque
byte array and accessed the si_addr field using unsafe (and
wrong on 386 and arm!) pointer arithmetic. Since si_addr is
the only field we use from this union, this replaces the
opaque byte array with an explicit declaration of the si_addr
field and accesses it directly.
Austin Clements [Wed, 19 Nov 2014 19:24:41 +0000 (14:24 -0500)]
[dev.cc] runtime: decode power64 branch instructions the way the CPU does
Previously, this used the top 8 bits of an instruction as a
sort-of opcode and ignored the top two bits of the relative
PC. This worked because these jumps are always negative and
never big enough for the top two bits of the relative PC (also
the bottom 2 bits of the sort-of opcode) to be anything other
than 0b11, but the code is confusing because it doesn't match
the actual structure of the instruction.
Instead, use the real 6 bit opcode and use all 24 bits of
relative PC.
LGTM=rsc
R=rsc, dave
CC=golang-codereviews
https://golang.org/cl/179960043
In go1.3, I type enter at the prompt and the program exits.
With the CL being rolled back, the program wedges at the
prompt.
««« original CL description
syscall: SysProcAttr job control changes
Making the child's process group the foreground process group and
placing the child in a specific process group involves co-ordination
between the parent and child that must be done post-fork but pre-exec.
Austin Clements [Wed, 19 Nov 2014 16:30:58 +0000 (11:30 -0500)]
[dev.cc] runtime: allow more address bits in lfstack on Power64
Previously, lfstack assumed Linux limited user space addresses
to 43 bits on Power64 based on a paper from 2001. It turns
out the limit is now 46 bits, so lfstack was truncating
pointers.
Raise the limit to 48 bits (for some future proofing and to
make it match amd64) and add a self-test that will fail in a
useful way if ever unpack(pack(x)) != x.
With this change, dev.cc passes all.bash on power64le.
Austin Clements [Tue, 18 Nov 2014 20:18:52 +0000 (15:18 -0500)]
[dev.cc] 9a: make RET a synonym for RETURN; use "g" instead of "R30"
Previously, 9a was the only assembler that had a different
name for RET, causing unnecessary friction in simple files
that otherwise assembled on all architectures. Add RET so
these work on 9a.
This also renames "R30" to "g" to avoid unintentionally
clobbering g in assembly code. This parallels a change made
to 5a.
Russ Cox [Tue, 18 Nov 2014 17:07:50 +0000 (12:07 -0500)]
[dev.cc] runtime: generate GOOS- and GOARCH-specific files with go generate
Eventually I'd like almost everything cmd/dist generates
to be done with 'go generate' and checked in, to simplify
the bootstrap process. The only thing cmd/dist really needs
to do is write things like the current experiment info and
the current version.
This is a first step toward that. It replaces the _NaCl etc
constants with generated ones goos_nacl, goos_darwin,
goarch_386, and so on.
LGTM=dave, austin
R=austin, dave, bradfitz
CC=golang-codereviews, iant, r
https://golang.org/cl/174290043
Austin Clements [Mon, 17 Nov 2014 19:44:41 +0000 (14:44 -0500)]
cmd/pprof: fix EOF handling when getting function source
getFunctionSource gathers five lines of "margin" around every
requested sample line. However, if this margin went past the
end of the source file, getFunctionSource would encounter an
io.EOF error and abort with this error, resulting in listings
like
(pprof) list main.main
ROUTINE ======================== main.main in ...
0 8.33s (flat, cum) 99.17% of Total
Error: EOF
(pprof)
Modify the error handling in getFunctionSource so io.EOF is
always considered non-fatal. If it reaches EOF, it simply
returns the lines it has.
Russ Cox [Mon, 17 Nov 2014 01:52:45 +0000 (20:52 -0500)]
debug/goobj: move to cmd/internal/goobj
debug/goobj is not ready to be published but it is
needed for the various binary-reading commands.
Move to cmd/internal/goobj.
(The Go 1.3 release branch deleted it, but that's not
an option anymore due to the command dependencies.
The API is still not vetted nor terribly well designed.)
David du Colombier [Sun, 16 Nov 2014 21:55:07 +0000 (22:55 +0100)]
[dev.cc] cmd/8g: work around "out of fixed registers" on Plan 9
This change works around the "out of fixed registers"
issue with the Plan 9 C compiler on 386, introduced by
the Bits change to uint64 in CL 169060043.
The purpose of this CL is to be able to properly
follow the conversion of the Plan 9 runtime to Go
on the Plan 9 builders.
This CL could be reverted once the Go compilers will
be converted to Go.
Thanks to Nick Owens for investigating this issue.
Russ Cox [Sun, 16 Nov 2014 21:44:45 +0000 (16:44 -0500)]
runtime: fix sudog leak
The SudoG used to sit on the stack, so it was cheap to allocated
and didn't need to be cleaned up when finished.
For the conversion to Go, we had to move sudog off the stack
for a few reasons, so we added a cache of recently used sudogs
to keep allocation cheap. But we didn't add any of the necessary
cleanup before adding a SudoG to the new cache, and so the cached
SudoGs had stale pointers inside them that have caused all sorts
of awful, hard to debug problems.
CL 155760043 made sure SudoG.elem is cleaned up.
CL 150520043 made sure SudoG.selectdone is cleaned up.
This CL makes sure SudoG.next, SudoG.prev, and SudoG.waitlink
are cleaned up. I should have done this when I did the other two
fields; instead I wasted a week tracking down a leak they caused.
A dangling SudoG.waitlink can point into a sudogcache list that
has been "forgotten" in order to let the GC collect it, but that
dangling .waitlink keeps the list from being collected.
And then the list holding the SudoG with the dangling waitlink
can find itself in the same situation, and so on. We end up
with lists of lists of unusable SudoGs that are still linked into
the object graph and never collected (given the right mix of
non-trivial selects and non-channel synchronization).
Dave Cheney [Sat, 15 Nov 2014 02:27:05 +0000 (13:27 +1100)]
[dev.cc] runtime: fix _sfloat thunk
* _sfloat dispatches to runtime._sfloat2 with the Go calling convention, so the seecond argument is a [15]uint32, not a *[15]uint32.
* adjust _sfloat2 to return the new pc in 68(R13) as expected.
David du Colombier [Fri, 14 Nov 2014 21:57:33 +0000 (22:57 +0100)]
[dev.cc] liblink: fix warnings on Plan 9
warning: src/liblink/asm9.c:501 set and not used: bflag
warning: src/liblink/list9.c:259 format mismatch .5lux INT, arg 4
warning: src/liblink/list9.c:261 format mismatch .5lux INT, arg 3
warning: src/liblink/list9.c:319 more arguments than format VLONG
warning: src/liblink/obj9.c:222 set and not used: autoffset
Austin Clements [Fri, 14 Nov 2014 18:58:31 +0000 (13:58 -0500)]
[dev.power64] 6g,9g: formatters for Prog and Addr details
The pretty printers for these make it hard to understand
what's actually in the fields of these structures. These
"ugly printers" show exactly what's in each field, which can
be useful for understanding and debugging code.
Austin Clements [Fri, 14 Nov 2014 17:08:46 +0000 (12:08 -0500)]
[dev.power64] liblink: generate dnames[5689] for D_* constants
This is more complicated than the other enums because the D_*
enums are full of explicit initializers and repeated values.
This tries its best. (This will get much cleaner once we
tease these constants apart better.)
Austin Clements [Fri, 14 Nov 2014 16:56:31 +0000 (11:56 -0500)]
[dev.power64] 5g,6g,8g,9g: debug prints for regopt pass 6 and paint2
Theses were very helpful in understanding the regions and
register selection when porting regopt to 9g. Add them to the
other compilers (and improve 9g's successor debug print).
Austin Clements [Thu, 13 Nov 2014 18:34:20 +0000 (13:34 -0500)]
[dev.power64] 6g,8g: remove unnecessary and incorrect reg use scanning
Previously, the 6g and 8g registerizers scanned for used
registers beyond the end of a region being considered for
registerization. This ancient artifact was copied from the C
compilers, where it was probably necessary to track implicitly
used registers. In the Go compilers it's harmless (because it
can only over-restrict the set of available registers), but no
longer necessary because the Go compilers correctly track
register use/set information. The consequences of this extra
scan were (at least) that 1) we would not consider allocating
the AX register if there was a deferproc call in the future
because deferproc uses AX as a return register, so we see the
use of AX, but don't track that AX is set by the CALL, and 2)
we could not consider allocating the DX register if there was
a MUL in the future because MUL implicitly sets DX and (thanks
to an abuse of copyu in this code) we would also consider DX
used.
This commit fixes these problems by nuking this code.
Russ Cox [Wed, 12 Nov 2014 19:54:31 +0000 (14:54 -0500)]
[dev.cc] runtime: delete scalararg, ptrarg; rename onM to systemstack
Scalararg and ptrarg are not "signal safe".
Go code filling them out can be interrupted by a signal,
and then the signal handler runs, and if it also ends up
in Go code that uses scalararg or ptrarg, now the old
values have been smashed.
For the pieces of code that do need to run in a signal handler,
we introduced onM_signalok, which is really just onM
except that the _signalok is meant to convey that the caller
asserts that scalarg and ptrarg will be restored to their old
values after the call (instead of the usual behavior, zeroing them).
Scalararg and ptrarg are also untyped and therefore error-prone.
Go code can always pass a closure instead of using scalararg
and ptrarg; they were only really necessary for C code.
And there's no more C code.
For all these reasons, delete scalararg and ptrarg, converting
the few remaining references to use closures.
Once those are gone, there is no need for a distinction between
onM and onM_signalok, so replace both with a single function
equivalent to the current onM_signalok (that is, it can be called
on any of the curg, g0, and gsignal stacks).
The name onM and the phrase 'm stack' are misnomers,
because on most system an M has two system stacks:
the main thread stack and the signal handling stack.
Correct the misnomer by naming the replacement function systemstack.
Fix a few references to "M stack" in code.
The main motivation for this change is to eliminate scalararg/ptrarg.
Rick and I have already seen them cause problems because
the calling sequence m.ptrarg[0] = p is a heap pointer assignment,
so it gets a write barrier. The write barrier also uses onM, so it has
all the same problems as if it were being invoked by a signal handler.
We worked around this by saving and restoring the old values
and by calling onM_signalok, but there's no point in keeping this nice
home for bugs around any longer.
This CL also changes funcline to return the file name as a result
instead of filling in a passed-in *string. (The *string signature is
left over from when the code was written in and called from C.)
That's arguably an unrelated change, except that once I had done
the ptrarg/scalararg/onM cleanup I started getting false positives
about the *string argument escaping (not allowed in package runtime).
The compiler is wrong, but the easiest fix is to write the code like
Go code instead of like C code. I am a bit worried that the compiler
is wrong because of some use of uninitialized memory in the escape
analysis. If that's the reason, it will go away when we convert the
compiler to Go. (And if not, we'll debug it the next time.)
Russ Cox [Tue, 11 Nov 2014 22:05:55 +0000 (17:05 -0500)]
[dev.cc] runtime: convert signal handlers from C to Go
This code overused macros and could not be
converted automatically. Instead a new sigctxt
type had to be defined for each os/arch combination,
with a common (implicit) interface used by the
arch-specific signal handler code.
[This CL is part of the removal of C code from package runtime.
See golang.org/s/dev.cc for an overview.]
Robert Griesemer [Tue, 11 Nov 2014 21:19:47 +0000 (13:19 -0800)]
spec: method selectors don't auto-deref named pointer types
Language clarification.
The existing rules for selector expressions imply
automatic dereferencing of pointers to struct fields.
They also implied automatic dereferencing of selectors
denoting methods. In almost all cases, such automatic
dereferencing does indeed take place for methods but the
reason is not the selector rules but the fact that method
sets include both methods with T and *T receivers; so for
a *T actual receiver, a method expecting a formal T
receiver, also accepts a *T (and the invocation or method
value expression is the reason for the auto-derefering).
However, the rules as stated so far implied that even in
case of a variable p of named pointer type P, a selector
expression p.f would always be shorthand for (*p).f. This
is true for field selectors f, but cannot be true for
method selectors since a named pointer type always has an
empty method set.
Named pointer types may never appear as anonymous field
types (and method receivers, for that matter), so this
only applies to variables declared of a named pointer
type. This is exceedingly rare and perhaps shouldn't be
permitted in the first place (but we cannot change that).
Amended the selector rules to make auto-deref of values
of named pointer types an exception to the general rules
and added corresponding examples with explanations.
Both gc and gccgo have a bug where they do auto-deref
pointers of named types in method selectors where they
should not:
See http://play.golang.org/p/c6VhjcIVdM , line 45.
Fixes #5769.
Fixes #8989.
LGTM=r, rsc
R=r, rsc, iant, ken
CC=golang-codereviews
https://golang.org/cl/168790043