Commit c23f057f authored by Sergei Azovskov's avatar Sergei Azovskov Committed by Simon Marlow

Mark code related symbols as @function not @object

This diff is a part of the bigger project which goal is to improve
common profiling tools support (perf) for GHC binaries.

A similar job was already done and reverted in the past:


`Perf` and similar tools build in memory symbol table from the .symtab
section of the ELF file to display human-readable function names instead
of the addresses in the output. `Perf` uses only two types of symbols:
`@function` and `@notype` but GHC is not capable to produce any
`@function` symbols so the `perf` output is pretty useless (All the
haskell symbols that you can see in `perf` now are `@notype` internal
symbols extracted by mistake/hack).

The changes:
 * mark code related symbols as @function
 * small hack to mark InfoTable symbols as code if TABLES_NEXT_TO_CODE is true

 * The perf symbolization support is not complete after this patch but
   I'm working on the second patch.
 * Constructor symbols are not supported. To fix that we can issue extra
   local symbols which mark code sections as code and will be only used
   for debug.

Test Plan:
any additional ideas?

Perf output on stock ghc 8.4.1:
     9.78%  FibbSlow  FibbSlow            [.] ckY_info
     9.59%  FibbSlow  FibbSlow            [.] cjqd_info
     7.17%  FibbSlow  FibbSlow            [.] c3sg_info
     6.62%  FibbSlow  FibbSlow            [.] c1X_info
     5.32%  FibbSlow  FibbSlow            [.] cjsX_info
     4.18%  FibbSlow  FibbSlow            [.] s3rN_info
     3.82%  FibbSlow  FibbSlow            [.] c2m_info
     3.68%  FibbSlow  FibbSlow            [.] cjlJ_info
     3.26%  FibbSlow  FibbSlow            [.] c3sb_info
     3.19%  FibbSlow  FibbSlow            [.] cjPQ_info
     3.05%  FibbSlow  FibbSlow            [.] cjQd_info
     2.97%  FibbSlow  FibbSlow            [.] cjAB_info
     2.78%  FibbSlow  FibbSlow            [.] cjzP_info
     2.40%  FibbSlow  FibbSlow            [.] cjOS_info
     2.38%  FibbSlow  FibbSlow            [.] s3rK_info
     2.27%  FibbSlow  FibbSlow            [.] cjq0_info
     2.18%  FibbSlow  FibbSlow            [.] cKQ_info
     2.13%  FibbSlow  FibbSlow            [.] cjSl_info
     1.99%  FibbSlow  FibbSlow            [.] s3rL_info
     1.98%  FibbSlow  FibbSlow            [.] c2cC_info
     1.80%  FibbSlow  FibbSlow            [.] s3rO_info
     1.37%  FibbSlow  FibbSlow            [.] c2f2_info

Perf output on patched ghc:
     7.97%  FibbSlow  FibbSlow            [.] c3rM_info
     6.75%  FibbSlow  FibbSlow            [.] 0x000000000032cfa8
     6.63%  FibbSlow  FibbSlow            [.] cifA_info
     4.98%  FibbSlow  FibbSlow            [.] integerzmgmp_GHCziIntegerziType_eqIntegerzh_info
     4.55%  FibbSlow  FibbSlow            [.] chXn_info
     4.52%  FibbSlow  FibbSlow            [.] c3rH_info
     4.45%  FibbSlow  FibbSlow            [.] chZB_info
     4.04%  FibbSlow  FibbSlow            [.] Main_fibbzuslow_info
     4.03%  FibbSlow  FibbSlow            [.] stg_ap_0_fast
     3.76%  FibbSlow  FibbSlow            [.] chXA_info
     3.67%  FibbSlow  FibbSlow            [.] cifu_info
     3.25%  FibbSlow  FibbSlow            [.] ci4r_info
     2.64%  FibbSlow  FibbSlow            [.] s3rf_info
     2.42%  FibbSlow  FibbSlow            [.] s3rg_info
     2.39%  FibbSlow  FibbSlow            [.] integerzmgmp_GHCziIntegerziType_eqInteger_info
     2.25%  FibbSlow  FibbSlow            [.] integerzmgmp_GHCziIntegerziType_minusInteger_info
     2.17%  FibbSlow  FibbSlow            [.] ghczmprim_GHCziClasses_zeze_info
     2.09%  FibbSlow  FibbSlow            [.] cicc_info
     2.03%  FibbSlow  FibbSlow            [.] 0x0000000000331e15
     2.02%  FibbSlow  FibbSlow            [.] s3ri_info
     1.91%  FibbSlow  FibbSlow            [.] 0x0000000000331bb8
     1.89%  FibbSlow  FibbSlow            [.] ci4N_info

Reviewers: simonmar, niteria, bgamari, goldfire

Reviewed By: simonmar, bgamari

Subscribers: lelf, rwbarton, thomie, carter

GHC Trac Issues: #15501

Differential Revision:
parent 64c54fff
......@@ -104,7 +104,9 @@ module CLabel (
-- * Conversions
toClosureLbl, toSlowEntryLbl, toEntryLbl, toInfoLbl, hasHaskellName,
) where
#include "HsVersions.h"
......@@ -621,6 +623,19 @@ isSomeRODataLabel (IdLabel _ _ BlockInfoTable) = True
isSomeRODataLabel (CmmLabel _ _ CmmInfo) = True
isSomeRODataLabel _lbl = False
-- | Whether label is points to some kind of info table
isInfoTableLabel :: CLabel -> Bool
isInfoTableLabel (IdLabel _ _ InfoTable) = True
isInfoTableLabel (IdLabel _ _ LocalInfoTable) = True
isInfoTableLabel (IdLabel _ _ ConInfoTable) = True
isInfoTableLabel (IdLabel _ _ BlockInfoTable) = True
isInfoTableLabel _ = False
-- | Whether label is points to constructor info table
isConInfoTableLabel :: CLabel -> Bool
isConInfoTableLabel (IdLabel _ _ ConInfoTable) = True
isConInfoTableLabel _ = False
-- | Get the label size field from a ForeignLabel
foreignLabelStdcallInfo :: CLabel -> Maybe Int
foreignLabelStdcallInfo (ForeignLabel _ info _ _) = info
......@@ -170,16 +170,76 @@ pprGloblDecl lbl
| not (externallyVisibleCLabel lbl) = empty
| otherwise = text ".globl " <> ppr lbl
pprTypeAndSizeDecl :: CLabel -> SDoc
pprTypeAndSizeDecl lbl
pprLabelType' :: DynFlags -> CLabel -> SDoc
pprLabelType' dflags lbl =
if isCFunctionLabel lbl || functionOkInfoTable then
text "@function"
text "@object"
NOTE: This is a bit hacky.
With the `tablesNextToCode` info tables look like this:
<info table data>
<info table code>
So actually info table label points exactly to the code and we can mark
the label as @function. (This is required to make perf and potentially other
tools to work on Haskell binaries).
This usually works well but it can cause issues with a linker.
A linker uses different algorithms for the relocation depending on
the symbol type.For some reason, a linker will generate JUMP_SLOT relocation
when constructor info table is referenced from a data section.
This only happens with static constructor call so
we mark _con_info symbols as `@object` to avoid the issue with relocations.
@SimonMarlow hack explanation:
"The reasoning goes like this:
* The danger when we mark a symbol as `@function` is that the linker will
redirect it to point to the PLT and use a `JUMP_SLOT` relocation when
the symbol refers to something outside the current shared object.
A PLT / JUMP_SLOT reference only works for symbols that we jump to, not
for symbols representing data,, nor for info table symbol references which
we expect to point directly to the info table.
* GHC generates code that might refer to any info table symbol from the text
segment, but that's OK, because those will be explicit GOT references
generated by the code generator.
* When we refer to info tables from the data segment, it's either
* a FUN_STATIC/THUNK_STATIC local to this module
* a `con_info` that could be from anywhere
So, the only info table symbols that we might refer to from the data segment
of another shared object are `con_info` symbols, so those are the ones we
need to exclude from getting the @function treatment.
A good place to check for more
Another possible hack is to create an extra local function symbol for
every code-like thing to give the needed information for to the tools
but mess up with the relocation.
functionOkInfoTable = tablesNextToCode dflags &&
isInfoTableLabel lbl && not (isConInfoTableLabel lbl)
pprTypeDecl :: CLabel -> SDoc
pprTypeDecl lbl
= sdocWithPlatform $ \platform ->
if osElfTarget (platformOS platform) && externallyVisibleCLabel lbl
then text ".type " <> ppr lbl <> ptext (sLit ", @object")
sdocWithDynFlags $ \df ->
text ".type " <> ppr lbl <> ptext (sLit ", ") <> pprLabelType' df lbl
else empty
pprLabel :: CLabel -> SDoc
pprLabel lbl = pprGloblDecl lbl
$$ pprTypeAndSizeDecl lbl
$$ pprTypeDecl lbl
$$ (ppr lbl <> char ':')
......@@ -1346,4 +1406,3 @@ pprFormatOpOpCoerce name format1 format2 op1 op2
pprCondInstr :: LitString -> Cond -> SDoc -> SDoc
pprCondInstr name cond arg
= hcat [ char '\t', ptext name, pprCond cond, space, arg]
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment