Keith Randall [Sat, 1 Jun 2013 03:58:31 +0000 (20:58 -0700)]
runtime: do hashmap grow work during reads.
Before this change, grow work was done only
during map writes to ensure multithreaded safety.
This can lead to maps remaining in a partially
grown state for a long time, potentially forever.
This change allows grow work to happen during reads,
which will lead to grow work finishing sooner, making
the resulting map smaller and faster.
Grow work is not done in parallel. Reads can
happen in parallel while grow work is happening.
Keith Randall [Sat, 1 Jun 2013 03:43:33 +0000 (20:43 -0700)]
runtime/gc: Run garbage collector on g0 stack
instead of regular g stack. We do this so that the g stack
we're currently running on is no longer changing. Cuts
the root set down a bit (g0 stacks are not scanned, and
we don't need to scan gc's internal state). Also an
enabler for copyable stacks.
Carl Shapiro [Fri, 31 May 2013 20:34:57 +0000 (13:34 -0700)]
cmd/5g, cmd/6g, cmd/8g: provide embedded trampolines with argument size information
An embedded trampoline is a function that exists to marshal
a receiver of type *S to a receiver of type *T when T is an
embedded field in S.
Embedded trampolines are generated by a special path through
the compiler and are not subject to the general analysis and
annotation done to functions. Their effects must be provided
explicitly.
Alberto García Hierro [Fri, 31 May 2013 18:33:36 +0000 (11:33 -0700)]
cmd/go: Add support for including C++ files in packages
* Add a CXXFiles field to Package, which includes .cc, .cpp and .cxx files.
* CXXFiles are compiled using g++, which can be overridden using the CXX environment variable.
* Include .hh, .hpp and .hxx files in HFiles.
* Add support for CPPFLAGS (used for both C and C++) and CXXFLAGS (used only for C++) in cgo directive.
* Changed pkg-config cgo directive to modify CPPFLAGS rather than CFLAGS, so both C and C++ files get any flag returned by pkg-config --cflags.
Fixes #1476.
R=iant, r
CC=bradfitz, gobot, golang-dev, iant, minux.ma, remyoudompheng, seb.binet
https://golang.org/cl/8248043
Dmitriy Vyukov [Fri, 31 May 2013 06:58:50 +0000 (10:58 +0400)]
runtime: fix heap coalescing bug introduced in cl/9802043
mheap.map become a pointer, so nelem(h->map) returns 1 rather than the map size.
As the result coalescing with subsequent spans does not happen.
Dmitriy Vyukov [Fri, 31 May 2013 06:42:30 +0000 (10:42 +0400)]
runtime: introduce helper persistentalloc() function
It is a caching wrapper around SysAlloc() that can allocate small chunks.
Use it for symtab allocations. Reduces number of symtab walks from 4 to 3
(reduces buildfuncs time from 10ms to 7.5ms on a large binary,
reduces initial heap size by 680K on the same binary).
Also can be used for type info allocation, itab allocation.
There are also several places in GC where we do the same thing,
they can be changed to use persistentalloc().
Also can be used in FixAlloc, because each instance of FixAlloc allocates
in 128K regions, which is too eager.
Reincarnation of committed and rolled back https://golang.org/cl/9805043
The latent bugs that it revealed are fixed:
https://golang.org/cl/9837049
https://golang.org/cl/9778048
Keith Randall [Fri, 31 May 2013 04:32:20 +0000 (21:32 -0700)]
runtime: set MSpan.limit properly for large spans.
Then use the limit to make sure MHeap_LookupMaybe & inlined
copies don't return a span if the pointer is beyond the limit.
Use this fact to optimize all call sites.
Rob Pike [Thu, 30 May 2013 15:28:08 +0000 (11:28 -0400)]
testing: quantize AllocsPerRun
As the code now says:
We are forced to return a float64 because the API is silly, but do
the division as integers so we can ask if AllocsPerRun()==1
instead of AllocsPerRun()<2.
Carl Shapiro [Thu, 30 May 2013 00:16:57 +0000 (17:16 -0700)]
cmd/ld, runtime: emit pointer maps for nosplits identified by the linker
A nosplits was assumed to have no argument information and no
pointer map. However, nosplits created by the linker often
have both. This change uses the pointer map size as an
alternate source of argument size when processing a nosplit.
In addition, the symbol table construction pointer map size
and argument size consistency check is strengthened. If a
nptrs is greater than 0 it must be equal to the number of
argument words.
Shenghou Ma [Wed, 29 May 2013 19:03:52 +0000 (03:03 +0800)]
cmd/5a, cmd/dist, runtime: support m/g in the assembler, drop support for R9/R10
to avoid unintentionally clobber R9/R10.
Thanks Lucio for the suggestion.
PS: yes, this could be considered a big change (but not an API change), but
as it turns out even temporarily changes R9/R10 in user code is unsafe and
leads to very hard to diagnose problems later, better to disable using R9/R10
when the user first uses it.
See CL 6300043 and CL 6305100 for two problems caused by misusing R9/R10.
Rob Pike [Wed, 29 May 2013 15:29:29 +0000 (11:29 -0400)]
fmt: change evalutation of indexed arg to match docs
The old code put the index before the period in the precision;
it should be after so it's always before the star, as documented.
A little trickier to do in one pass but compensated for by more
tests and catching a couple of other error cases.
John Shahid [Wed, 29 May 2013 15:21:32 +0000 (11:21 -0400)]
crypto/tls: Check all certificates in the path.
Currently we only check the leaf node's issuer against the list of
distinguished names in the server's CertificateRequest message. This
will fail if the client certiciate has more than one certificate in
the path and the leaf node issuer isn't in the list of distinguished
names, but the issuer's issuer was in the distinguished names.
Dmitriy Vyukov [Wed, 29 May 2013 07:49:45 +0000 (11:49 +0400)]
runtime: make notetsleep() return false if timeout happens
This is needed for preemptive scheduler, because during
stoptheworld we want to wait with timeout and re-preempt
M's on timeout.
Carl Shapiro [Wed, 29 May 2013 00:59:10 +0000 (17:59 -0700)]
cmd/5l, cmd/6l, cmd/8l, cmd/gc, runtime: generate and use bitmaps of argument pointer locations
With this change the compiler emits a bitmap for each function
covering its stack frame arguments area. If an argument word
is known to contain a pointer, a bit is set. The garbage
collector reads this information when scanning the stack by
frames and uses it to ignores locations known to not contain a
pointer.
Dmitriy Vyukov [Tue, 28 May 2013 18:14:47 +0000 (22:14 +0400)]
runtime: make mheap statically allocated again
This depends on: 9791044: runtime: allocate page table lazily
Once page table is moved out of heap, the heap becomes small.
This removes unnecessary dereferences during heap access.
No logical changes.
Dmitriy Vyukov [Tue, 28 May 2013 18:04:34 +0000 (22:04 +0400)]
runtime: allocate page table lazily
This removes the 256MB memory allocation at startup,
which conflicts with ulimit.
Also will allow to eliminate an unnecessary memory dereference in GC,
because the page table is usually mapped at known address.
Update #5049.
Update #5236.
Dmitriy Vyukov [Tue, 28 May 2013 17:09:27 +0000 (21:09 +0400)]
os/exec: fix test hang
Currently the test closes random files descriptors,
which leads to hang (in particular if netpoll fd is closed).
Try to open only fd 3, since the parent process expects it to be fd 3 anyway.
Fixes #5571.
R=golang-dev, r
CC=golang-dev
https://golang.org/cl/9778048
Dmitriy Vyukov [Tue, 28 May 2013 15:17:47 +0000 (19:17 +0400)]
runtime: fix heap corruption during GC
The 'n' variable is used during rescan initiation in GC_END case,
but it's overwritten with chan capacity in GC_CHAN case.
As the result rescan is done with the wrong object size.
Fixes #5554.
««« original CL description
runtime: introduce helper persistentalloc() function
It is a caching wrapper around SysAlloc() that can allocate small chunks.
Use it for symtab allocations. Reduces number of symtab walks from 4 to 3
(reduces buildfuncs time from 10ms to 7.5ms on a large binary,
reduces initial heap size by 680K on the same binary).
Also can be used for type info allocation, itab allocation.
There are also several places in GC where we do the same thing,
they can be changed to use persistentalloc().
Also can be used in FixAlloc, because each instance of FixAlloc allocates
in 128K regions, which is too eager.
Dmitriy Vyukov [Tue, 28 May 2013 06:47:35 +0000 (10:47 +0400)]
runtime: introduce helper persistentalloc() function
It is a caching wrapper around SysAlloc() that can allocate small chunks.
Use it for symtab allocations. Reduces number of symtab walks from 4 to 3
(reduces buildfuncs time from 10ms to 7.5ms on a large binary,
reduces initial heap size by 680K on the same binary).
Also can be used for type info allocation, itab allocation.
There are also several places in GC where we do the same thing,
they can be changed to use persistentalloc().
Also can be used in FixAlloc, because each instance of FixAlloc allocates
in 128K regions, which is too eager.
Jan Ziak [Mon, 27 May 2013 06:11:59 +0000 (08:11 +0200)]
runtime: flag static variables as no-pointers
Variables in data sections of 32-bit executables interfere with
garbage collector's ability to free objects and/or unnecessarily
slow down the garbage collector.
This changeset moves some static variables to .noptr sections.
'files' in symtab.c is now allocated dynamically.
Rob Pike [Fri, 24 May 2013 22:49:26 +0000 (15:49 -0700)]
fmt.Printf: introduce notation for random access to arguments.
This text is added to doc.go:
Explicit argument indexes:
In Printf, Sprintf, and Fprintf, the default behavior is for each
formatting verb to format successive arguments passed in the call.
However, the notation [n] immediately before the verb indicates that the
nth one-indexed argument is to be formatted instead. The same notation
before a '*' for a width or precision selects the argument index holding
the value. After processing a bracketed expression [n], arguments n+1,
n+2, etc. will be processed unless otherwise directed.
For example,
fmt.Sprintf("%[2]d %[1]d\n", 11, 22)
will yield "22, 11", while
fmt.Sprintf("%[3]*[2].*[1]f", 12.0, 2, 6),
equivalent to
fmt.Sprintf("%6.2f", 12.0),
will yield " 12.00". Because an explicit index affects subsequent verbs,
this notation can be used to print the same values multiple times
by resetting the index for the first argument to be repeated:
fmt.Sprintf("%d %d %#[1]x %#x", 16, 17)
will yield "16 17 0x10 0x11".
The notation chosen differs from that in C, but I believe it's easier to read
and to remember (we're indexing the arguments), and compatibility with
C's printf was never a strong goal anyway.
While we're here, change the word "field" to "arg" or "argument" in the
code; it was being misused and was confusing.
David du Colombier [Fri, 24 May 2013 20:55:19 +0000 (13:55 -0700)]
build: fix make.rc on Plan 9
Set $status as null to prevent rc from exiting
on the last --no-banner argument checking when
used with rc -e. It allows all.rc to not exit
before executing run.rc
Dmitriy Vyukov [Fri, 24 May 2013 14:35:48 +0000 (18:35 +0400)]
log/syslog: fix deadlock in test
The problem was that server handlers block on done<-,
the goroutine that reads from done blocks on count<-,
and the main goroutine that is supposed to read from count
waits for server handlers to exit.
Fixes #5547.
Dominik Honnef [Thu, 23 May 2013 17:19:03 +0000 (10:19 -0700)]
misc/emacs: Do not modify kill ring when programmatically deleting text
Operations like gofmt and go-remove-unused-imports delete entire
lines of text. Previously this put them on the kill-ring,
negatively affecting user experience.
Daniel Morsing [Thu, 23 May 2013 16:29:19 +0000 (18:29 +0200)]
io: Prioritize WriterTos over ReaderFroms in Copy.
This only affects calls where both ReaderFrom and WriterTo are implemented. WriterTo can issue one large write, while ReaderFrom must Read until EOF, potentially reallocating when out of memory. With one large Write, the Writer only needs to allocate once.
This also helps in ioutil.Discard since we can avoid copying memory when the Reader implements WriterTo.
Dmitriy Vyukov [Wed, 22 May 2013 19:04:46 +0000 (23:04 +0400)]
runtime: properly synchronize GC and finalizer goroutine
This is needed for preemptive scheduler, because the goroutine
can be preempted at surprising points.
Dmitriy Vyukov [Wed, 22 May 2013 18:57:47 +0000 (22:57 +0400)]
runtime: detect deadlocks in programs using cgo
When cgo is used, runtime creates an additional M to handle callbacks on threads not created by Go.
This effectively disabled deadlock detection, which is a right thing, because Go program can be blocked
and only serve callbacks on external threads.
This also disables deadlock detection under race detector, because it happens to use cgo.
With this change the additional M is created lazily on first cgo call. So deadlock detector
works for programs that import "C", "net" or "net/http/pprof" but do not use them in fact.
Also fixes deadlock detector under race detector.
It should be fine to create the M later, because C code can not call into Go before first cgo call,
because C code does not know when Go initialization has completed. So a Go program need to call into C
first either to create an external thread, or notify a thread created in global ctor that Go
initialization has completed.
Fixes #4973.
Fixes #5475.
Dmitriy Vyukov [Wed, 22 May 2013 18:22:57 +0000 (22:22 +0400)]
runtime: fix and speedup malloc stats
Currently per-sizeclass stats are lost for destroyed MCache's. This patch fixes this.
Also, only update mstats.heap_alloc on heap operations, because that's the only
stat that needs to be promptly updated. Everything else needs to be up-to-date only in ReadMemStats().
Dmitriy Vyukov [Wed, 22 May 2013 09:29:17 +0000 (13:29 +0400)]
runtime: simplify MCache
The nlistmin/size thresholds are copied from tcmalloc,
but are unnecesary for Go malloc. We do not do explicit
frees into MCache. For sparse cases when we do (mainly hashmap),
simpler logic will do.
Brad Fitzpatrick [Tue, 21 May 2013 22:51:49 +0000 (15:51 -0700)]
bufio: reuse Writer buffers after Flush
A bufio.Writer.Flush marks the usual end of a Writer's
life. Recycle its internal buffer on those explicit flushes,
but not on normal, as-needed internal flushes.
benchmark old ns/op new ns/op delta
BenchmarkWriterEmpty 1959 727 -62.89%
benchmark old allocs new allocs delta
BenchmarkWriterEmpty 2 1 -50.00%
benchmark old bytes new bytes delta
BenchmarkWriterEmpty 4215 83 -98.03%
Adam Langley [Tue, 21 May 2013 14:47:31 +0000 (10:47 -0400)]
crypto/tls: don't send NPN extension if NextProtos is not set.
This isn't clearly a bug on Go's part, but it triggers a bug in Firefox
which means that crypto/tls and net/http cannot be wired up together
unless NextProtos includes "http/1.1". When net/http sets up the
tls.Config, it does this and so works fine. But anyone setting up the
tls.Config themselves will hit the Firefox bug.
Fixes #5445.
R=golang-dev, bradfitz, r
CC=golang-dev
https://golang.org/cl/9539045