| Commit message (Collapse) | Author | Age |
|
|
|
| |
svn:r13594
|
|
|
|
|
|
|
| |
Fix all but 2 DOCDOC items; defer many XXX020s (particularly those where fixing them would fix no bugs at the risk of introducing some bugs).
svn:r13529
|
|
|
|
|
|
|
| |
Re-tune mempool parametes based on testing on peacetime: use smaller chuncks, free them a little more aggressively, and try very hard to concentrate allocations on fuller chunks. Also, lots of new documentation.
svn:r13484
|
|
|
|
|
|
|
| |
Add a bunch more code documentation; change the interface of fetch_var_cell_from_buf() so it takes the current link protocol into account and can't get confused by weird command bytes on v1 connections.
svn:r13430
|
|
|
|
| |
svn:r13412
|
|
|
|
|
|
|
| |
Add more documentation; change the behavior of read_to_buf_tls to be more consistent. Note a longstanding problem with current read/write interfaces.
svn:r13407
|
|
|
|
|
|
|
| |
Fix a bunch of DOCDOC items; document the --quiet flag; refactor a couple of XXXX020 items.
svn:r13405
|
|
|
|
|
|
|
| |
Fix/downgrade some more XXX020s.
svn:r13397
|
|
|
|
|
|
|
| |
Add a --disable-buffer-freelists configure argument to build without buffer RAM chunk freelists, so we can benchmark the impact of this.
svn:r13121
|
|
|
|
|
|
|
| |
fix whitespace
svn:r13035
|
|
|
|
|
|
|
| |
Make HTTP-processing code in buffers.c faster: do not pull up 50K of text into the first chunk if the headers end earlier than that.
svn:r13006
|
|
|
|
|
|
|
| |
Incomplete code to avoid doing needless pull-ups on HTTP. Also, use memstr instead of strstr to find the content-length header.
svn:r13005
|
|
|
|
|
|
|
| |
Try even harder not to allocate chunks bigger than MAX_CHUNK_ALLOC.
svn:r13003
|
|
|
|
|
|
|
| |
Remove need for buf_pullup in fetch_line_from_buf().
svn:r13002
|
|
|
|
|
|
|
| |
Exciting new entries in buf_dump_freelist_sizes() to make sure our freelist sizes are reasonable.
svn:r12996
|
|
|
|
|
|
|
| |
Fix windows compile error
svn:r12995
|
|
|
|
|
|
|
| |
And clean up an assertion failure.
svn:r12993
|
|
|
|
|
|
|
| |
Apparently loops work better when you actually increment/decrement the loop variable. Who knew?
svn:r12992
|
|
|
|
|
|
|
| |
Actually trim freelists.
svn:r12991
|
|
|
|
|
|
|
| |
Fix a nasty infinite loop in flush_buf[_tls].
svn:r12990
|
|
|
|
|
|
|
| |
More tweaks on 0-byte TLS writes.
svn:r12989
|
|
|
|
|
|
|
| |
Oops. flush_buf_tls can request more than the requested number of bytes. When that happens, do not let the size_t sz wrap around.
svn:r12988
|
|
|
|
|
|
|
| |
Peacetime managed to run overnight in paranoia mode. Disabling paranoia, so that buffers.c can be a little speedy again.
svn:r12987
|
|
|
|
|
|
|
| |
Fix in flush_buf_tls: it is okay to flush an empty buffer, since we may have a partial TLS record pending.
svn:r12984
|
|
|
|
|
|
|
| |
Here, have some terribly clever new buffer code. It uses a mbuf-like strategy rather than a ring buffer strategy, so it should require far far less extra memory to hold any given amount of data. Also, it avoids access patterns like x=malloc(1024);x=realloc(x,1048576);x=realloc(x,1024);append_to_freelist(x) that might have been contributing to memory fragmentation. I've tested it out a little on peacetime, and it seems to work so far. If you want to benchmark it for speed, make sure to remove the #define PARANOIA; #define NOINLINE macros at the head of the module.
svn:r12983
|
|
|
|
| |
svn:r12786
|
|
|
|
|
|
|
| |
more fixes into rend-spec.txt.
svn:r12715
|
|
|
|
|
|
|
| |
Make http://torproject urls into https://torproject urls. Patch from steve.
svn:r12708
|
|
|
|
|
|
|
| |
patch from steve: replace more tor.eff.org instances with torproject.org
svn:r12707
|
|
|
|
| |
svn:r12608
|
|
|
|
| |
svn:r12607
|
|
|
|
|
|
|
| |
Make VERSIONS variable-length.
svn:r12391
|
|
|
|
|
|
|
| |
Initial code for variable-length cells. CERT and VERSIONS need to use them.
svn:r12390
|
|
|
|
|
|
|
|
|
|
|
|
| |
edge_connection_t: want_onehop if it must attach to a circuit with
only one hop (e.g. for the current tunnelled connections that use
begin_dir), and use_begindir if we mean to use a BEGIN_DIR relay
command to establish the stream rather than the normal BEGIN. Now
we can make anonymized begin_dir connections for (e.g.) more secure
hidden service posting and fetching.
svn:r12244
|
|
|
|
|
|
| |
weasel less
svn:r11834
|
|
|
|
|
|
|
| |
Make controllers accept LF as well as CRLF. Update spec to reflect this. Remove now-dead code. Make controller warning about v0 protocol more accurate.
svn:r11299
|
|
|
|
|
|
|
| |
possible fix for bug 473: always set buf->cur on buf_remove_from_front.
svn:r11004
|
|
|
|
|
|
| |
mallocs() be inefficient, and we havent needed it in a long long time.
svn:r10994
|
|
|
|
|
|
| |
around for an empty buffer that could go on the freelist. This wants profiling to make sure that performance doesnt suffer.
svn:r10993
|
|
|
|
|
|
|
| |
Weasel noticed that many buffers spend their time with empty 4k, 8k, and 16k memory chunks. Thus, be more aggressive about putting empty chunks on the freelist, regardless of their high water marks. (Also, run buffer_shrink_freelist on the 8k-chunk freelist.)
svn:r10953
|
|
|
|
|
|
|
|
|
| |
Fix warnings on platforms where rlim values can be signed.
Add an 8k buffer freelist.
svn:r10948
|
|
|
|
|
|
|
| |
When dumping memory usage, list bytes used in buffer memory free-lists.
svn:r10947
|
|
|
|
| |
svn:r10854
|
|
|
|
| |
svn:r10851
|
|
|
|
|
|
|
| |
Sun CC likes to give warnings for the do { } while(0) construction for making statement-like macros. Define STMT_BEGIN/STMT_END macros that do the right thing, and use them everywhere.
svn:r10645
|
|
|
|
|
|
|
| |
Remove some deadcode that seems unlikely to return
svn:r10509
|
|
|
|
|
|
|
| |
When freeing slack buffers, free the buffers at the end of the freelist first.
svn:r10313
|
|
|
|
|
|
|
| |
Enable (and cope with) more GCC 4.2 warnings.
svn:r10196
|
|
|
|
|
|
|
| |
Make buffer unit tests pass again
svn:r10158
|
|
|
|
| |
svn:r10152
|