[PATCH 2/5] dirmngr: hkp: avoid race condition when some hosts die

[Werner Koch]

On Tue,  1 Nov 2016 01:33, dkg at fifthhorseman.net said:

> Multiple threads may write to hosttable[x]->dead while
> select_random_host() is running.  For example, a housekeeping thread

Nope.  Please look at the code:

  for (idx=0, tblsize=0; (pidx = table[idx]) != -1; idx++)
    if (hosttable[pidx] && !hosttable[pidx]->dead)
      tblsize++;
  if (!tblsize)
    return -1; /* No hosts.  */

  tbl = xtrymalloc (tblsize * sizeof *tbl);
  if (!tbl)
    return -1;
  for (idx=0, tblsize=0; (pidx = table[idx]) != -1; idx++)
    if (hosttable[pidx] && !hosttable[pidx]->dead)
      tbl[tblsize++] = pidx;

  if (tblsize == 1)  /* Save a get_uint_nonce.  */
    pidx = tbl[0];
  else
    pidx = tbl[get_uint_nonce () % tblsize];

and explain where you see a syscall which may switch to another thread.
Neither xtrymalloc nor get_unit_nonce are such syscalls or supposed to
use nPth wrapper syscalls.

I agree that the use of get_unit_nonce is a bit fragile because we can't
immediatly see how it is implemented and thus would be better to get its
value at the start of the function.


Shalom-Salam,

   Werner

-- 
Die Gedanken sind frei.  Ausnahmen regelt ein Bundesgesetz.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 194 bytes
Desc: not available
URL: </pipermail/attachments/20161101/472d93c2/attachment-0001.sig>