public inbox for binutils-cvs@sourceware.org
 help / color / mirror / Atom feed
* [binutils-gdb] libsframe: fix sframe_find_fre for pltN entries
@ 2023-06-09 18:48 Indu Bhagat
  0 siblings, 0 replies; 2+ messages in thread
From: Indu Bhagat @ 2023-06-09 18:48 UTC (permalink / raw)
  To: bfd-cvs, gdb-cvs

https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;h=937c461e41b866fc26d0c4ac6973ee3923267aab

commit 937c461e41b866fc26d0c4ac6973ee3923267aab
Author: Indu Bhagat <indu.bhagat@oracle.com>
Date:   Fri Jun 9 11:14:05 2023 -0700

    libsframe: fix sframe_find_fre for pltN entries
    
    To find SFrame stack trace information from an FDE of type
    SFRAME_FDE_TYPE_PCMASK, sframe_find_fre () was doing an operation
    like,
      (start_ip_offset & 0xff) >= (pc & 0xff), etc.
    
    This is buggy and needs correction.  The mask 0xff should be 0xf (to
    work for a pltN entry of size say, 16 bytes).
    
    At this time, the size of the pltN entry is implicitly assumed to be 16
    bytes by libsframe.  In next version of the SFrame format, we can encode
    this information explicitly in the SFrame FDE.
    
    For now, we should fix the code to at least behave correctly for the
    generated code and the generated SFrame stack trace information for the
    pltN entries on x86_64.
    
    libsframe/
            * sframe.c (sframe_find_fre): Correct the bitmask used for
            SFrame FDEs of type SFRAME_FDE_TYPE_PCMASK.

Diff:
---
 libsframe/sframe.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/libsframe/sframe.c b/libsframe/sframe.c
index a5f4a7f6519..7308a45ce88 100644
--- a/libsframe/sframe.c
+++ b/libsframe/sframe.c
@@ -1066,7 +1066,7 @@ sframe_find_fre (sframe_decoder_ctx *ctx, int32_t pc,
   /* FIXME - the bitmask should be picked per ABI or encoded in the format
      somehow. For AMD64, the pltN entry stub is 16 bytes. */
   if (fde_type == SFRAME_FDE_TYPE_PCMASK)
-    bitmask = 0xff;
+    bitmask = 0xf;
 
   fres = ctx->sfd_fres + fdep->sfde_func_start_fre_off;
   func_start_addr = fdep->sfde_func_start_address;

^ permalink raw reply	[flat|nested] 2+ messages in thread

* [binutils-gdb] libsframe: fix sframe_find_fre for pltN entries
@ 2023-06-29 18:13 Indu Bhagat
  0 siblings, 0 replies; 2+ messages in thread
From: Indu Bhagat @ 2023-06-29 18:13 UTC (permalink / raw)
  To: bfd-cvs, gdb-cvs

https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;h=3169b734cf07ec8800436b2c5298897aa993d2be

commit 3169b734cf07ec8800436b2c5298897aa993d2be
Author: Indu Bhagat <indu.bhagat@oracle.com>
Date:   Thu Jun 29 11:03:32 2023 -0700

    libsframe: fix sframe_find_fre for pltN entries
    
    For a toy application on x86_64, for example, following is the SFrame
    stack trace information for the 3 pltN entries of 16 bytes each:
    
       func idx [1]: pc = 0x401030, size = 48 bytes
       STARTPC[m]      CFA       FP        RA
       0000000000000000  sp+8      u         u
       000000000000000b  sp+16     u         u
    
    The data in first column is the start_ip_offset.  Also note that the FDE
    is of type SFRAME_FDE_TYPE_PCMASK (denoted by the [m] on LHS).
    
    Where each pltN (note: excluding plt0 entry) entry looks like:
    
      401030: jmp    *0x2fca(%rip)
      401036: push   $0x0
      40103b: jmp    401020<_init+0x20>
    
      401040: jmp    *0x2fc2(%rip)
      401046: push   $0x1
      40104b: jmp    401020<_init+0x20>
    
      401050: jmp    *0x2fba(%rip)
      401056: push   $0x2
      40105b: jmp    401020<_init+0x20>
    
    Now, to find SFrame stack trace information from an FDE of type
    SFRAME_FDE_TYPE_PCMASK, sframe_find_fre () was doing an operation
    like,
      (start_ip_offset & 0xf) >= (pc & 0xf)
    
    This works for pltN entry of size, say, less than 16 bytes.  But if the
    pltN entries or similar code stubs (for which SFrame FDE of type
    SFRAME_FDE_TYPE_PCMASK may be used), evolve to be of size > 16 bytes,
    this will cease to work.
    
    To match the range covered by the SFrame FRE, one should instead perform
    a modulo operation.  The constant for the modulo operation must be the
    size of the pltN entry.  Further, this constant should ideally be
    encoded in the format, as it may be different for each ABI.
    
    In SFrame Version 2 of the format, we will move towards encoding it
    explicitly in the SFrame FDE.  For now, fix up the logic to at least
    move towards modulo operation.
    
    libsframe/
            * sframe.c (sframe_fre_check_range_p): New definition.
            (sframe_find_fre): Refactor a bit and use the new definition
            above.
    include/
            * sframe.h (SFRAME_FDE_TYPE_PCMASK): Update comment.
    libsframe/doc/
            * sframe-spec.texi: Fix the text for SFRAME_FDE_TYPE_PCMASK FDE
            type.

Diff:
---
 include/sframe.h               |  4 +--
 libsframe/doc/sframe-spec.texi |  6 ++--
 libsframe/sframe.c             | 78 ++++++++++++++++++++++++++++++------------
 3 files changed, 62 insertions(+), 26 deletions(-)

diff --git a/include/sframe.h b/include/sframe.h
index c3dbb3a4097..cdf275f69e4 100644
--- a/include/sframe.h
+++ b/include/sframe.h
@@ -117,8 +117,8 @@ extern "C"
 
 /* Unwinders perform a (PC >= FRE_START_ADDR) to look up a matching FRE.  */
 #define SFRAME_FDE_TYPE_PCINC   0
-/* Unwinders perform a (PC & FRE_START_ADDR_AS_MASK >= FRE_START_ADDR_AS_MASK)
-   to look up a matching FRE.  */
+/* Unwinders perform a (PC % REP_BLOCK_SIZE >= FRE_START_ADDR) to look up a
+   matching FRE.  */
 #define SFRAME_FDE_TYPE_PCMASK  1
 
 typedef struct sframe_preamble
diff --git a/libsframe/doc/sframe-spec.texi b/libsframe/doc/sframe-spec.texi
index 5155410bdc8..a37a6f91414 100644
--- a/libsframe/doc/sframe-spec.texi
+++ b/libsframe/doc/sframe-spec.texi
@@ -451,8 +451,10 @@ entries and trampolines.
 
 @item SFRAME_FDE_TYPE_PCMASK
 @tab 1 @tab  Unwinders perform a @*
-(PC & FRE_START_ADDR_AS_MASK >= FRE_START_ADDR_AS_MASK)
-to look up a matching FRE.
+(PC % REP_BLOCK_SIZE @*
+ >= FRE_START_ADDR)
+to look up a matching FRE.  REP_BLOCK_SIZE is the size in bytes of the
+repeating block of program instructions.
 
 @end multitable
 
diff --git a/libsframe/sframe.c b/libsframe/sframe.c
index 29142436b84..fd966cfffd4 100644
--- a/libsframe/sframe.c
+++ b/libsframe/sframe.c
@@ -362,6 +362,53 @@ sframe_decoder_get_funcdesc_at_index (sframe_decoder_ctx *ctx,
   return fdep;
 }
 
+/* Check whether for the given FDEP, the SFrame Frame Row Entry identified via
+   the START_IP_OFFSET and the END_IP_OFFSET, provides the stack trace
+   information for the PC.  */
+
+static bool
+sframe_fre_check_range_p (sframe_func_desc_entry *fdep,
+			  int32_t start_ip_offset, int32_t end_ip_offset,
+			  int32_t pc)
+{
+  int32_t start_ip, end_ip;
+  int32_t func_start_addr;
+  uint32_t rep_block_size;
+  uint32_t fde_type;
+  int32_t masked_pc;
+  bool mask_p;
+  bool ret;
+
+  ret = false;
+  /* FIXME - the rep_block_size should be encoded in the format somehow.  For
+     AMD64, each pltN entry stub in .plt is 16 bytes.  */
+  rep_block_size = 16;
+
+  if (!fdep)
+    return ret;
+
+  func_start_addr = fdep->sfde_func_start_address;
+  fde_type = sframe_get_fde_type (fdep);
+  mask_p = (fde_type == SFRAME_FDE_TYPE_PCMASK);
+
+  if (!mask_p)
+    {
+      start_ip = start_ip_offset + func_start_addr;
+      end_ip = end_ip_offset + func_start_addr;
+      ret = ((start_ip <= pc) && (end_ip >= pc));
+    }
+  else
+    {
+      /* For FDEs for repetitive pattern of insns, we need to return the FRE
+	 where pc % rep_block_size is between start_ip_offset and
+	 end_ip_offset.  */
+      masked_pc = pc % rep_block_size;
+      ret = ((start_ip_offset <= masked_pc) && (end_ip_offset >= masked_pc));
+    }
+
+  return ret;
+}
+
 static int
 flip_fre (char *fp, uint32_t fre_type, size_t *fre_size)
 {
@@ -1056,19 +1103,14 @@ sframe_find_fre (sframe_decoder_ctx *ctx, int32_t pc,
 {
   sframe_frame_row_entry cur_fre;
   sframe_func_desc_entry *fdep;
-  uint32_t fre_type, fde_type;
-  uint32_t end_ip_offset, i;
-  int32_t start_ip, end_ip;
+  uint32_t fre_type, fde_type, i;
+  int32_t start_ip_offset;
   int32_t func_start_addr;
+  int32_t end_ip_offset;
   const char *fres;
   size_t size = 0;
   int err = 0;
-  /* For regular FDEs (i.e. fde_type SFRAME_FDE_TYPE_PCINC),
-     where the start address in the FRE is an offset from start pc,
-     use a bitmask with all bits set so that none of the address bits are
-     ignored.  In this case, we need to return the FRE where
-     (PC >= FRE_START_ADDR) */
-  uint64_t bitmask = 0xffffffff;
+  bool mask_p;
 
   if ((ctx == NULL) || (frep == NULL))
     return sframe_set_errno (&err, SFRAME_ERR_INVAL);
@@ -1080,14 +1122,7 @@ sframe_find_fre (sframe_decoder_ctx *ctx, int32_t pc,
 
   fre_type = sframe_get_fre_type (fdep);
   fde_type = sframe_get_fde_type (fdep);
-
-  /* For FDEs for repetitive pattern of insns, we need to return the FRE
-     such that (PC & FRE_START_ADDR_AS_MASK >= FRE_START_ADDR_AS_MASK).
-     so, update the bitmask to the start address.  */
-  /* FIXME - the bitmask should be picked per ABI or encoded in the format
-     somehow. For AMD64, the pltN entry stub is 16 bytes. */
-  if (fde_type == SFRAME_FDE_TYPE_PCMASK)
-    bitmask = 0xf;
+  mask_p = (fde_type == SFRAME_FDE_TYPE_PCMASK);
 
   fres = ctx->sfd_fres + fdep->sfde_func_start_fre_off;
   func_start_addr = fdep->sfde_func_start_address;
@@ -1098,15 +1133,14 @@ sframe_find_fre (sframe_decoder_ctx *ctx, int32_t pc,
      if (err)
        return sframe_set_errno (&err, SFRAME_ERR_FRE_INVAL);
 
-     start_ip = func_start_addr + cur_fre.fre_start_addr;
+     start_ip_offset = cur_fre.fre_start_addr;
      end_ip_offset = sframe_fre_get_end_ip_offset (fdep, i, fres + size);
-     end_ip = func_start_addr + end_ip_offset;
 
-     if ((start_ip & bitmask) > (pc & bitmask))
+     /* First FRE's start_ip must be more than pc for regular SFrame FDEs.  */
+     if (i == 0 && !mask_p && (start_ip_offset + func_start_addr) > pc)
        return sframe_set_errno (&err, SFRAME_ERR_FRE_INVAL);
 
-     if (((start_ip & bitmask) <= (pc & bitmask))
-	 && (end_ip & bitmask) >= (pc & bitmask))
+     if (sframe_fre_check_range_p (fdep, start_ip_offset, end_ip_offset, pc))
        {
 	 sframe_frame_row_entry_copy (frep, &cur_fre);
 	 return 0;

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2023-06-29 18:13 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-06-09 18:48 [binutils-gdb] libsframe: fix sframe_find_fre for pltN entries Indu Bhagat
2023-06-29 18:13 Indu Bhagat

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).