JDK 1.2.2_06 on Solaris has this problem where VM can crash while in GC. Customer uses JDBC/Oracle's OCI driver to make database calls. A single thread then iterates on the results obtained through the query. While this thread is running and GC happens to kick in, we have seen VM crashes with segv. The stack is as below:

(/mnt/sparc/opt/SUNWspro/SC4.2/bin/dbx) where
current thread: t@12
=>[1] cswToPreviousFrame(0xe5491634, 0x0, 0x244, 0xe61fe020, 0x38ce4, 0xe5805228), at 0xef4d753c
  [2] rjswToPreviousFrame(0xef72ce30, 0xef6dd0ac, 0x1, 0x1, 0xef6dd0ac, 0xe54916b4), at 0xef4d7a40
  [3] referencesInStackDo(0xef729d40, 0xef513150, 0x0, 0x0, 0xe5491950, 0xef6dd0ac), at 0xef468680
  [4] processRootsWork(0xef70c530, 0xef6dd0ac, 0xe5491950, 0xe54917e0, 0x4c7ad0, 0x1), at 0xef44d484
  [5] parallelDoDeg(0xef70c2e8, 0x4c7ad0, 0xef44d310, 0xe54917e0, 0x0, 0xe54917e0), at 0xef4cf0c0
  [6] processRoots(0x1, 0xef6c1e64, 0xef5134cc, 0xef513524, 0x0, 0x0), at 0xef44d9dc
  [7] processStrongRoots(0x1, 0xef6c1e64, 0xef5134cc, 0xef513524, 0x0, 0x0), at 0xef44da40
  [8] processStrongRefsIntoGen(0xef6e5fe4, 0x1, 0xef6c1e64, 0x0, 0x2, 0x0), at 0xef5136a0
  [9] ssCollect(0xef6e64ec, 0x343b, 0xb237fbf4, 0x51eb851f, 0xef6dd0ac, 0x37b70), at 0xef51fdc4
  [10] timedGC(0x0, 0xef70c2e8, 0x0, 0x0, 0xef6dd0ac, 0x379b8), at 0xef511ef4
  [11] specificGC(0xef72d7e0, 0x0, 0x14, 0x0, 0xef6e5fb0, 0x1), at 0xef5121bc
  [12] lockedGC(0xef6e458c, 0xef6dd0ac, 0x14, 0xee641604, 0xef71f6b4, 0x1), at 0xef45129c
  [13] stopWorldAndDoCurThread(0x0, 0xee641604, 0xef70c2e8, 0xef72d7e0, 0xef6dd0ac, 0x35d50d), at 0xef44e0b8
  [14] GCThread_run(0xef6dd0ac, 0xef70c500, 0xef6e4438, 0xef70c510, 0x1, 0x2000), at 0xef44e684
  [15] _start(0xef6e5df8, 0x1000, 0xef711158, 0xe54, 0x1, 0xef6f6000), at 0xef4fe468
(/mnt/sparc/opt/SUNWspro/SC4.2/bin/dbx) examine 0xe5491634

To reprdouce this problme, we wrote a small testcase program and installed Oracle 8i. Please see comments section for more details on how to invoke and run the test.