This commit implements a new config option: 'DisableAllSwap'
This option probably only works properly when Tor is started as root.
We added two new functions: tor_mlockall() and tor_set_max_memlock().
tor_mlockall() attempts to mlock() all current and all future memory pages.
For tor_mlockall() to work properly we set the process rlimits for memory to
RLIM_INFINITY (and beyond) inside of tor_set_max_memlock().
We behave differently from mlockall() by only allowing tor_mlockall() to be
called one single time. All other calls will result in a return code of 1.
It is not possible to change DisableAllSwap while running.
A sample configuration item was added to the torrc.complete.in config file.
A new item in the man page for DisableAllSwap was added.
Thanks to Moxie Marlinspike and Chris Palmer for their feedback on this patch.
Please note that we make no guarantees about the quality of your OS and its
mlock/mlockall implementation. It is possible that this will do nothing at all.
It is also possible that you can ulimit the mlock properties of a given user
such that root is not required. This has not been extensively tested and is
unsupported. I have included some comments for possible ways we can handle
this on win32.
I don't think we actually use (or plan to use) strtok_r in a reentrant
way anywhere in our code, but would be nice not to have to think about
whether we're doing it.
This patch adds a function to determine whether we're in the main
thread, and changes control_event_logmsg() to return immediately if
we're in a subthread. This is necessary because otherwise we will
call connection_write_to_buf, which modifies non-locked data
structures.
Bugfix on 0.2.0.x; fix for at least one of the things currently
called "bug 977".
This matters because a cpuworker can close its socket when it
finishes. Cpuworker typically runs in another thread, so without a
lock here, we can have a race condition and get confused about how
many sockets are open. Possible fix for bug 939.
It seems that 64-bit Sparc Solaris demands 64-bit-aligned access to
uint64_t, but does not 64-bit-align the stack-allocated char array we
use for cpuworker tags. So this patch adds a set/get_uint64 pair, and
uses them to access the conn_id field in the tag.
svn:r18743
The subversion $Id$ fields made every commit force a rebuild of
whatever file got committed. They were not actually useful for
telling the version of Tor files in the wild.
svn:r17867
Make definition of tor_mutex_t go into compat.h, so that it is possible to inline mutexes in critical objects. Add init/uninit functions for mutexes allocated inside other structs.
svn:r16623
Make generic address manipulation functions work better. Switch address policy code to use tor_addr_t, so it can handle IPv6. That is a good place to start.
svn:r16178
Forward-port: I had apparently broken OSX and Freebsd by not initializing threading before we initialize the logging system. This patch should do so, and fix bug 671.
svn:r14430
On platforms using pthreads, allow a thread to acquire a lock it already holds. This is crucial for logging: otherwise any log message thrown from inside the logging process (especially from control.c) will deadlock. Win32 CriticalSections are already recursive. Bug spotted by nwf. Bugfix on 0.2.0.16-alpha. Backport candidate. I hope this is portable.
svn:r14406
Combine common code in set_max_file_descriptors(): all that varies from platform to platform in the no-getrlimit() case is the connection limit and the platform name.
svn:r14101
Fix all remaining shorten-64-to-32 errors in src/common. Some were genuine problems. Many were compatibility errors with libraries (openssl, zlib) that like predate size_t. Partial backport candidate.
svn:r13665
Apply patch from Sebastian Hahn: stop imposing an arbitrary maximum on the number of file descriptors used for busy servers. Bug reported by Olaf Selke.
svn:r13626
Change tor_addr_t to be a tagged union of in_addr and in6_addr, not of sockaddr_in and sockaddr_in6. It's hardly used in the main code as it is, but let's get it right before it gets popular.
svn:r12660
If setting our rlimit to rlim_max or cap fails, fall back to OPEN_FILES if defiled. This makes Tor run on OSX 10.5, while allowing OSX to mend its ways in the future.
svn:r12341
Patch from croup: rewrite the logic of get_next_token() to do the right thing with input that ends at weird places, or aligns with block boundaries after mmap. should fix bug 455. Needs fuzzing.
svn:r10847
Implement conditions in compat.c; switch windows to use "critical sections" instead of mutexes. Apparently, mutexes are for IPC and critical sections are for multithreaded.
svn:r10716
First bare stubs of ipv6 work: commit some (untested, hence doublessly broken) implementations of inet_ntop/pton for systems that lack them.
svn:r10326
[Backport candidate] On windows, open cached-routers with the sharing mode "FILE_SHARE_READ so that other processes can read it while Tor is running. (Reported by Janbar).
svn:r10148
Track the number of connection_t separately from the number of open sockets. It is already possible to have connections that do not count: resolving conns, for one. Once we move from socketpairs to linked conns, and once we do dns proxying, there will be lots of such connections.
svn:r9994
Apply patch from coderman: have posix subthreads mask out signals. This could prevent some kinds of crashes when subthreads try to handle SIGPIPEs and die in the attempt. Backport candidate.
svn:r9603
Move all struct-offset-manipulation macros into util.h, and use them consistently. Because there are days when "SUBTYPE_P(handle, subtype, _base)" is just easier to read and write than "(basetp*)(((handle) - STRUCT_OFFSET(subtype, _base))".
svn:r9592
Handle errors on opening cached-routers* more uniformly and sanely: log not-found errors at level INFO, and all other errors at level WARN. Needs testing on win32.
svn:r9569
Removing the last DOCDOC comment hurt so much that I had to use Doxygen to identify undocumented macros and comments, and add 150 more DOCDOCs to point out where they were. Oops. Hey, kids! Fixing some of these could be your first Tor patch!
svn:r9477
Fix a couple of obvious bugs in tor_mmap_file on Windows: first, fix a boolean error when checking the return value of CreateFileMapping. Second, CreateFileMapping is documented to return NULL on failure.
svn:r9035
Try to compile with fewer warnings on irix64's MIPSpro compiler /
environment, which apparently believes that:
- off_t can be bigger than size_t.
- only mean kids assign things they do not subsequently inspect.
I don't try to fix the "error" that makes it say:
cc-3970 cc: WARNING File = main.c, Line = 1277
conversion from pointer to same-sized integral type (potential portability
problem)
uintptr_t sig = (uintptr_t)arg;
Because really, what can you do about a compiler that claims to be c99
but doesn't understand that void* x = NULL; uintptr_t y = (uintptr_t) x;
is safe?
svn:r8948
Add unit tests for tor_mmap_file(); make tor_mmap_t.size always be the size of the file (not the size of the mapping); add an extra argument to read_file_to_str() so it can return the size of the result string.
svn:r8762
Do not graciously increase the size to be mmaped if the current size already is
at a page_size boundary. This is important since if a file has a size of zero
and we mmap() it with length > 0, then accessing the mmaped memory area causes
a bus error. However, if we pass a length of 0 to mmap() it will return with -1
and things work from there.
svn:r8387
spawn_func fixes: have cpuworker_main and dnsworker_main confirm to the right interfaces [casting func to void* is icky]. Also, make pthread_create() build without warnings.
svn:r8327
- made configure link to required system dll's if building for win32
- added diffs for libevent 1.1b
- forced user to turn off eventdns if win32 is set
- cleaned up tor_mmap_file()_win32 (not sure if it's stable)
- cleaned up some warnings and typos
svn:r8322
possible. Cache at the server side independently from the TTL, to
prevent attackers from probing the server to see who has been asking
for what hostnames. (Hi, Dan Kaminski!)
Also, clean some whitespace.
svn:r6526
This is needed because Windows already has an err() that we
can't clobber. And we need to be able to make the log functions
a macro so we can print the function's name in the log entry.
svn:r6000
constraining ourselves to a max of 1024 connections. Now if there is no
getrlimit, just assume that we can handle the whole 15000 connections.
The better answer is to find a getrlimit equivalent on Windows, but hey,
one step at a time.
svn:r5854
