32 bit issues

[sans-sense]

building on 32 bit ubuntu 12.04 complains about unavailable values at intr.c:694 and generates core dump at dt_module.c:992.

I am not an expert, but the below patch works for me.

The code seems to declare gate_t as gate64 or gate32, but this block uses gate64 fields which are not available on my 32 bit machine. If I patch
--- a/driver/intr.c
+++ b/driver/intr.c
@@ -694,10 +694,14 @@ dtrace_write_idt_entry2(int vec)

info[0].vector = vec;
info[0].address = gate_offset(*val);

• // AN we do not have these for 32 bits
  +#ifndef __i386
  info[0].cs = gate_segment(*val);
  info[0].flags = val->dpl;
  if (val->type == GATE_INTERRUPT)
  info[0].flags |= 1 << 2;
  +#endif

It works.

I am also seeing coredumps from dt_module.c as dm_aslen seems to be 0, so when it enters the for loop, the first run with 0 gives problems, introducing a length check prevents this issue.

--- a/libdtrace/dt_module.c
+++ b/libdtrace/dt_module.c

@@ -992,9 +992,12 @@ asmap[i]->st_size);_/
}
} else {
Elf32_Sym *_asmap = dmp->dm_asmap;

•   for (i = 0; i < dmp->dm_aslen-1; i++) {
  

•    // AN no values for dmp->dm_aslen
  

•    if (dmp->dm_aslen > 0) {
  

•      for (i = 0; i < dmp->dm_aslen-1; i++) {
      asmap[i]->st_size = asmap[i+1]->st_value - asmap[i]->st_value;
  

•   }
  

•      }
  

•    }
  

I am running with the latest code and the last commit was commit 21b5732

[dtrace4linux]

Thanks Apurba for these two fixes. The second fix is suboptimal, so have
put in the same fix i did for x86_64; the first one I could not reproduce
on my 12.04 VM, but I have put it in anyhow.

Thanks for pointing out the latter issue - I was wondering why it was core
dumping during the usdt test.

On 10 December 2012 05:43, Apurba notifications@github.com wrote:

building on 32 bit ubuntu 12.04 complains about unavailable values at
intr.c:694 and generates core dump at dt_module.c:992.

I am not an expert, but the below patch works for me.

The code seems to declare gate_t as gate64 or gate32, but this block uses
gate64 fields which are not available on my 32 bit machine. If I patch
--- a/driver/intr.c
+++ b/driver/intr.c
@@ -694,10 +694,14 @@ dtrace_write_idt_entry2(int vec)

info[0].vector = vec;
info[0].address = gate_offset(*val);

• // AN we do not have these for 32 bits +#ifndef __i386 info[0].cs =
  gate_segment(*val); info[0].flags = val->dpl; if (val->type ==
  GATE_INTERRUPT) info[0].flags |= 1 << 2; +#endif +

It works.

I am also seeing coredumps from dt_module.c as dm_aslen seems to be 0, so
when it enters the for loop, the first run with 0 gives problems,
introducing a length check prevents this issue.

--- a/libdtrace/dt_module.c
+++ b/libdtrace/dt_module.c
@@ -992,9 +992,12 @@ asmap[i]->st_size);_/
}
} else {
Elf32_Sym *_asmap = dmp->dm_asmap;

• for (i = 0; i < dmp->dm_aslen-1; i++) {
• // AN no values for dmp->dm_aslen
• if (dmp->dm_aslen > 0) {
• for (i = 0; i < dmp->dm_aslen-1; i++) { asmap[i]->st_size =
  asmap[i+1]->st_value - asmap[i]->st_value;
• }
• }
• }

I am running with the latest code and the last commit was commit 21b573221b5732
21b573221b5732

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/44.