Bill,
I was under the impression from playing with 0.9.4 that the lock/unlock
variant of malloc/free had to do with memory being "locked" in actual storage
and not to be flushed to disk (virtual memory). This is to add an extra measure
of security in that in-core/in-memory info is harder to interrogate than disk
data.
- Jim Keohane
Bill Rebey wrote:
> This one's in the same ballpark as the others - having trouble cleaning up
> before thread exit.
>
> This time the crash was in ERR_get_state (...), and ultimately in a malloc.
> This happened on a SPARC Solaris box. As is typical of these bugs that I've
> been reporting, this happened after about 200,000 successful connections.
>
> The fact that there is a call to something called _malloc_unlocked is rather
> ominous - it sort of sounds like a single-threaded kind of malloc. Does
> anyone know of any magic link or compile stuff that I should be doing on
> Solaris to support MT that I'm not doing? Does anyone really KNOW about
> this _malloc_unlocked call? Is it REALLY a single-threaded call? Is this
> the source of my trouble?
>
> [1] t_delete(0x9, 0xef5a3138, 0x98, 0xe97e0, 0xda350, 0x150), at 0xef5464b8
> [2] realfree(0xe97e8, 0x9, 0xef5a9758, 0xef5a3138, 0xe97e0, 0x153), at
> 0xef546200
> [3] cleanfree(0x0, 0xef5a3138, 0xef5a96cc, 0xef5a974c, 0xef5a96e0, 0x0),
> at 0xef5469cc
> [4] _malloc_unlocked(0x150, 0xef5d5250, 0x0, 0xe9e08, 0x0, 0xef5a3138), at
> 0xef545c28
> [5] malloc(0x14c, 0x14c, 0x0, 0x0, 0x0, 0x0), at 0xef545b30
> [6] CRYPTO_malloc(0xcac6c, 0xb0a80, 0x2c0, 0xb0a80, 0x14c, 0x0), at
> 0x42628
> [7] ERR_get_state(0xb0a80, 0xef5d9c08, 0x4425, 0x0, 0x0, 0x5dde4), at
> 0x569ec
> [8] get_error_values(0x1, 0x0, 0x0, 0x0, 0x0, 0x0), at 0x561a0
> [9] 0xef786ca8(0xe5960, 0xef5aaa1c, 0xef5a6ae8, 0x27, 0x0, 0x61240a), at
> 0xef786ca7
> =>[10] __cAppThread::~__cAppThread(this = 0xe5960), line 304 in
> "SockTest.cpp"
> [11] __SLIP.DELETER__B(0xe5960, 0x1, 0x64, 0x26160, 0xe6ec8, 0xe5960), at
> 0x2b2b0
> [12] __cAppThread::ThreadMain(this = 0xe5960), line 329 in "SockTest.cpp"
> [13] ThreadRootStartingPoint(pThreadInstance = 0xe5960), line 74 in
> "cThread.cpp"
>
> Thanks, as usual, for the help!
>
> Bill Rebey
>
> ______________________________________________________________________
> OpenSSL Project http://www.openssl.org
> Development Mailing List [EMAIL PROTECTED]
> Automated List Manager [EMAIL PROTECTED]
--
Jim Keohane
Brigadier Consultant
LockStar, Inc.
1200 Wall Street West
Lyndhurst, NJ 07071
Tel: (201) 508-3231
Fax: (201) 508-3201
http://www.lockstar.com
"Anyone who considers protocol unimportant has never dealt with a cat"
-L.Long
______________________________________________________________________
OpenSSL Project http://www.openssl.org
Development Mailing List [EMAIL PROTECTED]
Automated List Manager [EMAIL PROTECTED]
