[v2] gdbserver: linux_low: elf_64_file_p cache results

From: Jon Ringle <jringle@gridpoint.com>

  From: Jon Ringle <jringle@gridpoint.com>

I was debugging a systemd service that would drop root privileges, and
found that I was unable to debug it properly because gdb could not resolve
any symbols. The gdbserver on my arm target would report:

  gdbserver: Corrupted shared library list: 0x0 != 0xe2822038

Prior to getting the above message, symbols could be resolved successfully.
I finally determined that the problem was that in linux_low.c in function
linux_qxfer_libraries_svr4() the is_elf64 variable was causing the wrong
lmo and ptr_size to get used. Here's that code snippet:

7136 static int
7137 linux_qxfer_libraries_svr4 (const char *annex, unsigned char *readbuf,
7138                             unsigned const char *writebuf,
7139                             CORE_ADDR offset, int len)
7140 {
7141   char *document;
7142   unsigned document_len;
7143   struct process_info_private *const priv = current_process ()->priv;
7144   char filename[PATH_MAX];
7145   int pid, is_elf64;
7146
7147   static const struct link_map_offsets lmo_32bit_offsets =
7148     {
7149       0,     /* r_version offset. */
7150       4,     /* r_debug.r_map offset.  */
7151       0,     /* l_addr offset in link_map.  */
7152       4,     /* l_name offset in link_map.  */
7153       8,     /* l_ld offset in link_map.  */
7154       12,    /* l_next offset in link_map.  */
7155       16     /* l_prev offset in link_map.  */
7156     };
7157
7158   static const struct link_map_offsets lmo_64bit_offsets =
7159     {
7160       0,     /* r_version offset. */
7161       8,     /* r_debug.r_map offset.  */
7162       0,     /* l_addr offset in link_map.  */
7163       8,     /* l_name offset in link_map.  */
7164       16,    /* l_ld offset in link_map.  */
7165       24,    /* l_next offset in link_map.  */
7166       32     /* l_prev offset in link_map.  */
7167     };
7168   const struct link_map_offsets *lmo;
7169   unsigned int machine;
7170   int ptr_size;
7171   CORE_ADDR lm_addr = 0, lm_prev = 0;
7172   int allocated = 1024;
7173   char *p;
7174   CORE_ADDR l_name, l_addr, l_ld, l_next, l_prev;
7175   int header_done = 0;
7176
7177   if (writebuf != NULL)
7178     return -2;
7179   if (readbuf == NULL)
7180     return -1;
7181
7182   pid = lwpid_of (current_thread);
7183   xsnprintf (filename, sizeof filename, "/proc/%d/exe", pid);
7184   is_elf64 = elf_64_file_p (filename, &machine);
7185   lmo = is_elf64 ? &lmo_64bit_offsets : &lmo_32bit_offsets;
7186   ptr_size = is_elf64 ? 8 : 4;

The problem lied in the fact that the function elf_64_file_p() (call on
line 7184), was returning -1 because it could not open the file (with errno
set to EACCESS). This was because the inferior's code was dropping root
privileges and no longer was able to read the file.

This patch implements a cache per file in the elf_64_file_p() function so
that it remembers the results of a previous query for the same filename.

Since it seems that c++ is now accepted in gdb, the cache was implemented
using std::map
---
 gdb/gdbserver/linux-low.c | 23 +++++++++++++++++------
 1 file changed, 17 insertions(+), 6 deletions(-)

[v2] gdbserver: linux_low: elf_64_file_p cache results

Commit Message

Comments

Patch