[PATCH v3 1/1] Don't rewind PC for GHC generated frames
Bartosz Nitka
niteria@gmail.com
Sun Feb 4 00:07:00 GMT 2018
GHC - the Haskell compiler generates code that violates one of
GDB's assumptions.
GDB assumes that the address in a frame was generated by the call
instruction and that it is the address right after the call
instruction (I'm paraphrasing the comment in get_frame_address_in_block).
So to get an address in the same block as the call instruction, one
has to substract 1. This is doubly beneficial because some functions
are "noreturn" and don't have further instructions after call,
so GDB would be looking at gibberish.
GHC generates completely different code. It uses jumps instead of call
and manages the stack itself. Furthermore every piece of code is
preceeded by some metadata called the Info Table. If we substract
from the program counter it ends up pointing to the Info Table,
which is undesirable.
GHC has a workaround for this [1] that works most of the time, it
basically lies in the DWARF data and extends the function one byte
backwards. That helps with making unwinding succeed most of the time,
but then the address is also used for looking up symbols and
they can't be resolved.
This change disables program counter rewinding for
compilation units generated by GHC.
Some additional context can be found here [2].
The impact.
Let's take an example from [2].
Here's the example Haskell program (fib.hs):
fib :: Int -> Int
fib 0 = 0
fib 1 = 1
fib n = fib (n-1) + fib (n-2)
main :: IO ()
main = print $ fib 20
If we run it with current GDB master we get:
(gdb) br fib.hs:3
Breakpoint 1 at 0x405588: file fib.hs, line 3.
(gdb) r
Breakpoint 1, Main_zdwfib_info () at fib.hs:3
3 fib 1 = 1
(gdb) bt
#0 Main_zdwfib_info () at fib.hs:3
#1 0x00000000004055c8 in Main_zdwfib_info () at fib.hs:4
#2 0x00000000004055c8 in Main_zdwfib_info () at fib.hs:4
#3 0x00000000004055c8 in Main_zdwfib_info () at fib.hs:4
#4 0x00000000004055c8 in Main_zdwfib_info () at fib.hs:4
#5 0x00000000004055c8 in Main_zdwfib_info () at fib.hs:4
#6 0x00000000004055c8 in Main_zdwfib_info () at fib.hs:4
#7 0x00000000004055c8 in Main_zdwfib_info () at fib.hs:4
#8 0x00000000004055c8 in Main_zdwfib_info () at fib.hs:4
#9 0x00000000004055c8 in Main_zdwfib_info () at fib.hs:4
#10 0x00000000004055c8 in Main_zdwfib_info () at fib.hs:4
#11 0x00000000004055c8 in Main_zdwfib_info () at fib.hs:4
#12 0x00000000004055c8 in Main_zdwfib_info () at fib.hs:4
#13 0x00000000004055c8 in Main_zdwfib_info () at fib.hs:4
#14 0x00000000004055c8 in Main_zdwfib_info () at fib.hs:4
#15 0x00000000004055c8 in Main_zdwfib_info () at fib.hs:4
#16 0x00000000004055c8 in Main_zdwfib_info () at fib.hs:4
#17 0x00000000004055c8 in Main_zdwfib_info () at fib.hs:4
#18 0x00000000004055c8 in Main_zdwfib_info () at fib.hs:4
#19 0x00000000004055c8 in Main_zdwfib_info () at fib.hs:4
#20 0x00000000004056d8 in Main_main2_info () at fib.hs:4
#21 0x000000000040bab0 in base_GHCziIOziHandleziText_zdwwriteBlocks_info () at libraries/base/GHC/IO/Handle/Text.hs:595
#22 0x0000000000488bf0 in ?? () at rts/Exception.cmm:332
Backtrace stopped: previous frame identical to this frame (corrupt stack?)
Here's an analogous session after this patch:
(gdb) br fib.hs:3
Breakpoint 1 at 0x405588: file fib.hs, line 3.
(gdb) r
Breakpoint 1, Main_zdwfib_info () at fib.hs:3
3 fib 1 = 1
(gdb) bt
#0 Main_zdwfib_info () at fib.hs:3
#1 Main_zdwfib_info () at fib.hs:4
#2 Main_zdwfib_info () at fib.hs:4
#3 Main_zdwfib_info () at fib.hs:4
#4 Main_zdwfib_info () at fib.hs:4
#5 Main_zdwfib_info () at fib.hs:4
#6 Main_zdwfib_info () at fib.hs:4
#7 Main_zdwfib_info () at fib.hs:4
#8 Main_zdwfib_info () at fib.hs:4
#9 Main_zdwfib_info () at fib.hs:4
#10 Main_zdwfib_info () at fib.hs:4
#11 Main_zdwfib_info () at fib.hs:4
#12 Main_zdwfib_info () at fib.hs:4
#13 Main_zdwfib_info () at fib.hs:4
#14 Main_zdwfib_info () at fib.hs:4
#15 Main_zdwfib_info () at fib.hs:4
#16 Main_zdwfib_info () at fib.hs:4
#17 Main_zdwfib_info () at fib.hs:4
#18 Main_zdwfib_info () at fib.hs:4
#19 Main_zdwfib_info () at fib.hs:4
#20 0x00000000004056d8 in Main_main2_info () at fib.hs:4
#21 base_GHCziIOziHandleziText_zdwwriteBlocks_info () at libraries/base/GHC/IO/Handle/Text.hs:582
#22 stg_catch_frame_info () at rts/Exception.cmm:370
#23 stg_stop_thread_info () at rts/StgStartup.cmm:42
#24 0x00000000004a62ab in c2MH_str ()
#25 0x00010000006c94c0 in ?? ()
#26 0x00000000006e5e60 in ?? ()
#27 0x00007fffffffdae8 in ?? ()
#28 0x00007fffffffdcb0 in ?? ()
#29 0x00000000006c94c0 in Main_main3_closure ()
#30 0x00007fffffffdbd0 in ?? ()
#31 0x0000000000405480 in ?? ()
#32 0x00007fffffffdcb0 in ?? ()
#33 0x0000000000000000 in ?? ()
There are a couple of things to note here.
First is that the unwinding got further, even a bit too far. It should
have ended on frame #23 on stg_stop_thread_info. That's the first thing
pushed on the stack when running a ligthweight thread. I'm not sure yet
how GHC is supposed to signal that it's the last thing on the stack and
if it's trying to do that.
Second thing is that we get the correct line number for frame #22 *and*
the symbol gets successfully resolved. Before the patch we were
substracting one, making the symbol lookup fail. I think that's also
what makes GDB give a cleaner output on frames #0 - #19.
Note that to get the same results as me, you have to compile the
compiler with debugging symbols as described on [2].
[1] https://phabricator.haskell.org/diffusion/GHC/browse/master/compiler/nativeGen/Dwarf/Types.hs;e9ae0cae9eb6a340473b339b5711ae76c6bdd045$399-417
[2] https://ghc.haskell.org/trac/ghc/wiki/DWARF
gdb/ChangeLog:
* dwarf2read.c (process_full_comp_unit): Populate
* producer_is_ghc.
* frame.c (get_frame_address_in_block): Don't rewind the program
counter for code generated by GHC.
* symtab.h (struct compunit_symtab): Add producer_is_ghc.
---
gdb/ChangeLog | 7 +++++++
gdb/dwarf2read.c | 4 ++++
gdb/frame.c | 9 ++++++++-
gdb/symtab.h | 3 +++
4 files changed, 22 insertions(+), 1 deletion(-)
diff --git a/gdb/ChangeLog b/gdb/ChangeLog
index 3ce980c8c3..6b35bf34b6 100644
--- a/gdb/ChangeLog
+++ b/gdb/ChangeLog
@@ -1,3 +1,10 @@
+2018-02-01 Bartosz Nitka <niteria@gmail.com>
+
+ * dwarf2read.c (process_full_comp_unit): Populate producer_is_ghc.
+ * frame.c (get_frame_address_in_block): Don't rewind the program
+ counter for code generated by GHC.
+ * symtab.h (struct compunit_symtab): Add producer_is_ghc.
+
2018-02-01 Yao Qi <yao.qi@linaro.org>
* arm-tdep.c (arm_record_data_proc_misc_ld_str): Rewrite it.
diff --git a/gdb/dwarf2read.c b/gdb/dwarf2read.c
index 51d0f39f75..2516c48741 100644
--- a/gdb/dwarf2read.c
+++ b/gdb/dwarf2read.c
@@ -10501,6 +10501,10 @@ process_full_comp_unit (struct dwarf2_per_cu_data *per_cu,
cust->epilogue_unwind_valid = 1;
cust->call_site_htab = cu->call_site_htab;
+
+ if (startswith (cu->producer,
+ "The Glorious Glasgow Haskell Compilation System"))
+ cust->producer_is_ghc = 1;
}
if (dwarf2_per_objfile->using_index)
diff --git a/gdb/frame.c b/gdb/frame.c
index 1384ecca4f..9ff0dcb130 100644
--- a/gdb/frame.c
+++ b/gdb/frame.c
@@ -2458,7 +2458,14 @@ get_frame_address_in_block (struct frame_info *this_frame)
&& (get_frame_type (this_frame) == NORMAL_FRAME
|| get_frame_type (this_frame) == TAILCALL_FRAME
|| get_frame_type (this_frame) == INLINE_FRAME))
- return pc - 1;
+ {
+ /* GHC intermixes metadata (info tables) with code, going back is
+ guaranteed to land us in the metadata. */
+ struct compunit_symtab *cust = find_pc_compunit_symtab (pc);
+ if (cust != NULL && cust->producer_is_ghc)
+ return pc;
+ return pc - 1;
+ }
return pc;
}
diff --git a/gdb/symtab.h b/gdb/symtab.h
index f9d52e7697..c164e5ba5f 100644
--- a/gdb/symtab.h
+++ b/gdb/symtab.h
@@ -1432,6 +1432,9 @@ struct compunit_symtab
instruction). This is supported by GCC since 4.5.0. */
unsigned int epilogue_unwind_valid : 1;
+ /* This CU was produced by Glasgow Haskell Compiler */
+ unsigned int producer_is_ghc : 1;
+
/* struct call_site entries for this compilation unit or NULL. */
htab_t call_site_htab;
--
2.14.1
More information about the Gdb-patches
mailing list