Tweak the code in trampoline generation that determines if a given
call branch will reach, changing the lower limit guard from "x <
-0x800000" to "x <= -0x800000". This is to resolve linking failures
when the computed displacement is exactly -0x800000, which results in
errors of the form
.../ld.gold: internal error in arm_branch_common, at ../../gold/arm.cc:4079
when using the Gold linker, and
...:(.text+0x...): relocation truncated to fit: R_ARM_CALL against `runtime.morestack_noctxt'
when using the bfd linker.
Fixes#59058.
Updates #59034.
Updates #58425.
Change-Id: I8a76986b38727df1b961654824c2af23f06b9fcf
Reviewed-on: https://go-review.googlesource.com/c/go/+/475957
Run-TryBot: Than McIntosh <thanm@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Cherry Mui <cherryyz@google.com>
(cherry picked from commit f26bf203ac)
Reviewed-on: https://go-review.googlesource.com/c/go/+/476935
This patch provides a fix for a problem linking large arm32 binaries
with external linking, specifically R_CALLARM relocations against
runtime.duff* routines being flagged by the external linker as not
reaching.
What appears to be happening in the bug in question is that the Go
linker and the external linker are using slightly different recipes to
decide whether a given R_CALLARM relocation will "fit" (e.g. will not
require a trampoline). The Go linker is taking into account the addend
on the call reloc (which for calls to runtime.duffcopy or
runtime.duffzero is nonzero), whereas the external linker appears to
be ignoring the addend.
Example to illustrate:
Addr Size Func
----- ----- -----
...
XYZ 1024 runtime.duffcopy
...
ABC ... mypackge.MyFunc
+ R0: R_CALLARM o=8 a=848 tgt=runtime.duffcopy<0>
Let's say that the distance between ABC (start address of
runtime.duffcopy) and XYZ (start of MyFunc) is just over the
architected 24-bit maximum displacement for an R_CALLARM (let's say
that ABC-XYZ is just over the architected limit by some small value,
say 36). Because we're calling into runtime.duffcopy at offset 848,
however, the relocation does in fact fit, but if the external linker
isn't taking into account the addend (assuming that all calls target
the first instruction of the called routine), then we'll get a
"doesn't fit" error from the linker.
To work around this problem, revise the ARM trampoline generation code
in the Go linker that computes the trampoline threshold to ignore the
addend on R_CALLARM relocations, so as to harmonize the two linkers.
Fixes#58502.
Updates #58428.
Updates #58425.
Change-Id: I56e580c05b7b47bbe8edf5532a1770bbd700fbe5
Reviewed-on: https://go-review.googlesource.com/c/go/+/469275
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Cherry Mui <cherryyz@google.com>
Run-TryBot: Than McIntosh <thanm@google.com>
(cherry picked from commit 0b5affb193)
Reviewed-on: https://go-review.googlesource.com/c/go/+/471598
It is now gone.
Change-Id: I59f68b324af706476695de2f291dd3aa5734e192
Reviewed-on: https://go-review.googlesource.com/c/go/+/351332
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
This is CL 314452, for ARM.
Fixes#30949.
Change-Id: Ib4e46a5bd11c698c4f8ea3bc4e7a605d7a538efc
Reviewed-on: https://go-review.googlesource.com/c/go/+/314455
Trust: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
The external linker uses R12. Do the same. We previously use R11,
the temp register in Go ABI. This does not really matter if the
caller is Go code, because all registers are clobbered at call.
But it the caller is C code, it may assume R11 live across a call.
Using R11 may clobber live value. On the callee side, R12 is not
an argument register in both Go and C calling convention.
Change-Id: I958c5dad52aa51bb282a7ad420f5423863e69c14
Reviewed-on: https://go-review.googlesource.com/c/go/+/314454
Trust: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
PPC64 needs to preserve bits when applying some relocations. DS form
relocations must preserve the lower two bits, and thus needs to inspect
the section data as it streams out.
Similarly, the overflow checking requires inspecting the primary
opcode to see if the value is sign or zero extended.
The existing PPC64 code no longer works as the slice returned by
(loader*).Data is cleared as we layout the symbol and process
relocations. This data is always the section undergoing relocation,
thus we can directly inspect the contents to preserve bits or
check for overflows.
Change-Id: I239211f7e5e96208673663b6553b3017adae7e01
Reviewed-on: https://go-review.googlesource.com/c/go/+/300555
Run-TryBot: Paul Murphy <murp@ibm.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Emmanuel Odeke <emmanuel@orijtech.com>
Currently, in the trampoline generation pass we expect packages
are laid out in dependency order, so a cross-package jump always
has a known target address so we can check if a trampoline is
needed. With linknames, there can be cycles in the package
dependency graph, making this algorithm no longer work. For them,
as the target address is unkown we conservatively generate a
trampoline. This may generate unnecessary trampolines (if the
packages turn out laid together), but package cycles are extremely
rare so this is fine.
Updates #44073.
Change-Id: I2dc2998edacbda27d726fc79452313a21d07787a
Reviewed-on: https://go-review.googlesource.com/c/go/+/292490
Trust: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
This reverts CL 252478.
Reason for revert: debug/Elfhdr has no Flags fields, some other CLs has removed it.
Change-Id: Ie199ac29f382c56aaf37a2e8338f2dafe6e79297
Reviewed-on: https://go-review.googlesource.com/c/go/+/265317
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Trust: Meng Zhuo <mzh@golangcn.org>
On ARM, for a JMP/CALL relocation, the instruction bytes is
encoded in Reloc.Add (issue #19811). I really hate it, but before
it is fixed we have to follow the rule and emit the right bits
from r.Add.
Fixes#40769.
Change-Id: I862e105408d344c5cc58ca9140d2e552e4364453
Reviewed-on: https://go-review.googlesource.com/c/go/+/248399
Reviewed-by: Jeremy Faller <jeremy@golang.org>
Reviewed-by: Joel Sing <joel@sing.id.au>
Reviewed-by: Than McIntosh <thanm@google.com>
Rename Reloc2 to Reloc, At2 to At, Aux2 to Aux.
Change-Id: Ic98d83c080e8cd80fbe1837c8f0aa134033508ce
Reviewed-on: https://go-review.googlesource.com/c/go/+/245578
Reviewed-by: Jeremy Faller <jeremy@golang.org>
We have Reloc and Reloc2. Reloc2 is the better approach and most
code uses Reloc2. There are still uses of Reloc. This CL migrates
them to Reloc2, and removes Reloc.
Change-Id: Id5f6a6019e1e044add682d05e70ebb1548ec58d9
Reviewed-on: https://go-review.googlesource.com/c/go/+/245577
Reviewed-by: Jeremy Faller <jeremy@golang.org>
We used to generate all external relocations in memory, then emit
the relocation records at a later pass. The data structures were
chosen so that it takes as little memory as possible. Now we just
stream out external relocations, and ExtReloc is just a local
variable. Change the data structure to avoid repeated read of
some fields. Also get rid of ExtRelocView, as it is no longer
necessary.
Change-Id: I40209bbe4387af231b29788125c3b4ebb0ff4a33
Reviewed-on: https://go-review.googlesource.com/c/go/+/245479
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Jeremy Faller <jeremy@golang.org>
Do them in the same CL so ARM's archreloc doesn't need to support
both streaming and non-streaming.
TODO: we haven't switched to using mmap to emit external
relocations on Windows.
Change-Id: Ica2ee89c03fc74839efd6b9e26c80585fcdce45c
Reviewed-on: https://go-review.googlesource.com/c/go/+/244357
Run-TryBot: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Jeremy Faller <jeremy@golang.org>
In CL 240399 we changed to precompute the size for ELF relocation
records and use mmap to write them, but we left architectures
where elfreloc1 write non-fixed number of bytes. This CL handles
those architectures. When a Go relocation will turn into multiple
ELF relocations, in relocsym we account this difference and add
it to the size calculation. So when emitting ELF relocations, we
know the number of ELF relocations to be emitted.
Change-Id: I6732ab674b442f4618405e5412a77f6e4a3315d7
Reviewed-on: https://go-review.googlesource.com/c/go/+/241079
Reviewed-by: Than McIntosh <thanm@google.com>
Reviewed-by: Jeremy Faller <jeremy@golang.org>
Now that we write ELF relocation records in mapped memory with
known sizes and offsets, we can write them in parallel.
Further speed up Asmb2 pass. Linking cmd/compile with external
linking,
Asmb2 141ms ± 4% 97ms ± 5% -30.98% (p=0.000 n=10+9)
Change-Id: I52c2b9230e90ed4421c21d7ef13a4f1e996f6054
Reviewed-on: https://go-review.googlesource.com/c/go/+/240400
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
The loader's SymbolBuilder Add*/Set* methods include a call to mark
the underlying symbol as reachable (as a convenience, so that callers
would not have to set it explicitly). This code was carried over from
the corresponding sym.Symbol methods; back in the sym.Symbol world
unreachable symbols were never removed from the AllSyms slice, hence
setting and checking reachability was a good deal more important.
With the advent of the loader and the new deadcode implementation,
there is less of a need for this sort of fallback, and in addition the
implicit attr setting introduces data races in the the loader if there
are SymbolBuilder Add*/Set* method calls in parallel threads, as well
as adding overhead to the methods.
This patch gets rid of the implicit reachability setting, and instead
marks reachability in CreateSymForUpdate, as well as adding a few
explicit SetAttrReachable calls where needed.
Change-Id: I029a0c5a4a24237826a7831f9cbe5180d44cbc40
Reviewed-on: https://go-review.googlesource.com/c/go/+/237678
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Jeremy Faller <jeremy@golang.org>
Merge conflicts are mostly recently changed nm/objdump output
format and its tests. Resolved easily (mostly just using the
format on master branch).
Change-Id: I99d8410a9a02947ecf027d9cae5762861562baf5
Lots of the architecture specific code for asmb() is very simimar. As
such, move to a common function.
Change-Id: Id1fd50ee7bfa1bc9978e3f42ad08914b04cd677b
Reviewed-on: https://go-review.googlesource.com/c/go/+/234683
Run-TryBot: Jeremy Faller <jeremy@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
The runtime needs to find the PC of the deferreturn call in a few
places. So for functions that have defer, we record the PC of
deferreturn call in its funcdata.
For very large binaries, the deferreturn call could be made
through a trampoline. The current code of finding deferreturn PC
fails in this case. This CL handles the trampoline as well.
Fixes#39049.
Change-Id: I929be54d6ae436f5294013793217dc2a35f080d4
Reviewed-on: https://go-review.googlesource.com/c/go/+/234105
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Jeremy Faller <jeremy@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
Minor renaming cleanup to get rid of a couple of old sym.Symbol
adddynrel helpers and rename the current crop of adddynrel2
methods/functions back to adddynrel.
Change-Id: I67e76decff84d603ef765f3b6a0cd78c7f3743ec
Reviewed-on: https://go-review.googlesource.com/c/go/+/233523
Run-TryBot: Than McIntosh <thanm@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Adds in support for remaining architectures to the linker's ELF asmb2
path, along with deleting most of the older sym.Symbol based code.
Change-Id: I67c96525db72b7d6dd3187cf2b9f6faddc296291
Reviewed-on: https://go-review.googlesource.com/c/go/+/233362
Reviewed-by: Cherry Zhang <cherryyz@google.com>
We use the new one everywhere now.
Change-Id: Ic9b1314e71e4666500cbf1689bb93839e040682a
Reviewed-on: https://go-review.googlesource.com/c/go/+/232982
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Jeremy Faller <jeremy@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
Now we no longer do loadlibfull on windows.
Change-Id: Ideb015597c28f27538bd50829e089ea728017162
Reviewed-on: https://go-review.googlesource.com/c/go/+/232979
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Jeremy Faller <jeremy@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
Now we no longer do loadlibfull on darwin.
While here, remove residual darwin/386 and darwin/arm code.
Change-Id: I6efdcd81baeeca29d1fe91c4fab0cc8241a58e2d
Reviewed-on: https://go-review.googlesource.com/c/go/+/232597
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Jeremy Faller <jeremy@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
Add code to the loader to enforce the invariant that there is only a
single level of 'outer' symbol nesting. That is, if outer(X) = Y, then
outer(Y) is always zero.
Revise foldSubSymbolOffset based on the new invariant, allowing it to
be inlined, and then fix the various "for s.Outer != nil" loops in the
linker to just use an "if" instead of a loop.
Change-Id: Ib895702bc6de52718248f09a5368b84cb2e0a3fa
Reviewed-on: https://go-review.googlesource.com/c/go/+/231137
Run-TryBot: Than McIntosh <thanm@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Currently, for external relocations, the ExtReloc structure
contains all the fields of the relocation. In fact, many of the
fields are the same with the original relocation. So, instead, we
can just use an index to reference the original relocation and
not expand the fields.
There is one place where we modify relocation type: changing
R_DWARFSECTREF to R_ADDR. Get away with it by changing
downstreams.
It also makes it easier to retrieve the reloc variant.
This reduces some allocation. Linking cmd/compile with external
linking,
name old alloc/op new alloc/op delta
Reloc_GC 34.1MB ± 0% 22.7MB ± 0% -33.30% (p=0.000 n=5+4)
Change-Id: Id08a89ed2aee705296886d3b95014b806a0d55cf
Reviewed-on: https://go-review.googlesource.com/c/go/+/231217
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
Reviewed-by: Jeremy Faller <jeremy@golang.org>
Adddynrel2 is a function pointer. In dynrelocsym we pass &r to
it, which will cause r to escape. Pass it by value instead.
Linking cmd/compile,
name old alloc/op new alloc/op delta
Dodata_GC 15.8MB ± 0% 5.9MB ± 0% -62.55% (p=0.008 n=5+5)
Change-Id: Ib86005d1026ebaca57777b27ead037e613585f44
Reviewed-on: https://go-review.googlesource.com/c/go/+/231045
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
Reviewed-by: Jeremy Faller <jeremy@golang.org>
Remove the "ElfSym" and "LocalElfSym" fields from sym.Symbol,
replacing uses with loader method calls as needed.
Change-Id: I3828f13203ece2bdc03eeb09ab37a5c94e21a726
Reviewed-on: https://go-review.googlesource.com/c/go/+/230462
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Make Wasm more like other architectures, writing data sections to
heap in Asmb instead of Asmb2. Then we can remove the
copy-on-write logic in applying relocations.
Change-Id: I26d5315ea9fba032fe4bdb9b5c7fe483611c4373
Reviewed-on: https://go-review.googlesource.com/c/go/+/230465
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
Reviewed-by: Jeremy Faller <jeremy@golang.org>
Add elf/ARM arch support for the new dodata() phase.
Change-Id: Iadd772b01036c6c5be95bcc6017f6c05d45a24c0
Reviewed-on: https://go-review.googlesource.com/c/go/+/229868
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Recreation of CL 228317.
The problem with that original CL was a late requested change,
reordering reloc and asmb, resulting in symbols having stale pointers to
their data. I've fixed this by preallocating the heap variable in OutBuf
for platforms w/o mmap.
Change-Id: Icdb392ac2c8d6518830f4c84cf422e78b8ab68c7
Reviewed-on: https://go-review.googlesource.com/c/go/+/228782
Run-TryBot: Jeremy Faller <jeremy@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
This removes all conditions and conditional code (that I could find)
that depended on darwin/arm.
Fixes#35439 (since that only happened on darwin/arm)
Fixes#37611.
Change-Id: Ia4c32a5a4368ed75231075832b0b5bfb1ad11986
Reviewed-on: https://go-review.googlesource.com/c/go/+/227198
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Than McIntosh
Cherry Mui
Paul E. Murphy
Meng Zhuo
Tobias Klauser
Jeremy Faller
Austin Clements