Commit Graph

21 Commits

Author SHA1 Message Date
Saleem Abdulrasool bb19a13c0b second pass over removal of Mutex and Condition
llvm-svn: 270024
2016-05-19 05:13:57 +00:00
Marianne Mailhot-Sarrasin 7a4eed280a Fixed MemoryCache L1 cache flush
Use the same method to find the cache line as in Read().

Differential Revision: http://reviews.llvm.org/D18050

llvm-svn: 263233
2016-03-11 13:50:10 +00:00
Jason Molenda a38312a9a4 Fix the L1 cache search in MemoryCache::Read to use the
stl upper_bound method instead of lower_bound - we were
failing to find some cached data in the L1 cache resulting
in extra memory read packets while stepping.

The bug with the existing code looked like this:
If the L1 cache has 8 bytes at address 0x1000 and 8 bytes
at address 0x2000 and we are searching for 4 bytes at 0x2004,
the use of lower_bound would return the end() of the container
and so we would incorrectly treat the memory as uncached.

(the L1 cache is memory seeded from debugserver in the T aka
questionmark packet, where debugserver will send up the stack
memory that likely contains the caller's stack pointer and 
frame pointer values.)

<rdar://problem/23869227> 

llvm-svn: 255421
2015-12-12 03:06:10 +00:00
Greg Clayton 358cf1ea30 Resubmitting 240466 after fixing the linux test suite failures.
A few extras were fixed

- Symbol::GetAddress() now returns an Address object, not a reference. There were places where people were accessing the address of a symbol when the symbol's value wasn't an address symbol. On MacOSX, undefined symbols have a value zero and some places where using the symbol's address and getting an absolute address of zero (since an Address object with no section and an m_offset whose value isn't LLDB_INVALID_ADDRESS is considered an absolute address). So fixing this required some changes to make sure people were getting what they expected. 
- Since some places want to access the address as a reference, I added a few new functions to symbol:
    Address &Symbol::GetAddressRef();
    const Address &Symbol::GetAddressRef() const;

Linux test suite passes just fine now.

<rdar://problem/21494354>

llvm-svn: 240702
2015-06-25 21:46:34 +00:00
Chaoren Lin 7a30608026 Revert "Reduced packet counts to the remote GDB server where possible."
This reverts commit 0cc0745ea9c68d7fdcadc9904cee3f13c96dae60.

Due to breakage on Linux build bot:

http://lab.llvm.org:8011/builders/lldb-x86_64-ubuntu-14.04-cmake/builds/3436

llvm-svn: 240371
2015-06-23 03:17:01 +00:00
Greg Clayton ffb2d44ab9 Reduced packet counts to the remote GDB server where possible.
We have been working on reducing the packet count that is sent between LLDB and the debugserver on MacOSX and iOS. Our approach to this was to reduce the packets required when debugging multiple threads. We currently make one qThreadStopInfoXXXX call (where XXXX is the thread ID in hex) per thread except the thread that stopped with a stop reply packet. In order to implement multiple thread infos in a single reply, we need to use structured data, which means JSON. The new jThreadsInfo packet will attempt to retrieve all thread infos in a single packet. The data is very similar to the stop reply packets, but packaged in JSON and uses JSON arrays where applicable. The JSON output looks like:


[
  { "tid":1580681,
    "metype":6,
    "medata":[2,0],
    "reason":"exception",
    "qaddr":140735118423168,
    "registers": {
      "0":"8000000000000000",
      "1":"0000000000000000",
      "2":"20fabf5fff7f0000",
      "3":"e8f8bf5fff7f0000",
      "4":"0100000000000000",
      "5":"d8f8bf5fff7f0000",
      "6":"b0f8bf5fff7f0000",
      "7":"20f4bf5fff7f0000",
      "8":"8000000000000000",
      "9":"61a8db78a61500db",
      "10":"3200000000000000",
      "11":"4602000000000000",
      "12":"0000000000000000",
      "13":"0000000000000000",
      "14":"0000000000000000",
      "15":"0000000000000000",
      "16":"960b000001000000",
      "17":"0202000000000000",
      "18":"2b00000000000000",
      "19":"0000000000000000",
      "20":"0000000000000000"},
    "memory":[
      {"address":140734799804592,"bytes":"c8f8bf5fff7f0000c9a59e8cff7f0000"},
      {"address":140734799804616,"bytes":"00000000000000000100000000000000"}
    ]
  }
]

It contains an array of dicitionaries with all of the key value pairs that are normally in the stop reply packet. Including the expedited registers. Notice that is also contains expedited memory in the "memory" key. Any values in this memory will get included in a new L1 cache in lldb_private::Process where if a memory read request is made and that memory request fits into one of the L1 memory cache blocks, it will use that memory data. If a memory request fails in the L1 cache, it will fall back to the L2 cache which is the same block sized caching we were using before these changes. This allows a process to expedite memory that you are likely to use and it reduces packet count. On MacOSX with debugserver, we expedite the frame pointer backchain for a thread (up to 256 entries) by reading 2 pointers worth of bytes at the frame pointer (for the previous FP and PC), and follow the backchain. Most backtraces on MacOSX and iOS now don't require us to read any memory!

We will try these packets out and if successful, we should port these to lldb-server in the near future. 

<rdar://problem/21494354>

llvm-svn: 240354
2015-06-22 23:12:45 +00:00
David Majnemer 8faf9370fa Clean-up warnings on Linux/GCC
llvm-svn: 217862
2014-09-16 06:34:29 +00:00
Jason Molenda f0340c9ab5 Add a new target.process.memory-cache-line-size to change the size of
lldb's internal memory cache chunks that are read from the remote
system.  For a remote connection that is especially slow, a user may
need to reduce it; reading a 512 byte chunk of memory whenever a
4-byte region is requested may not be the right decision in these
kinds of environments.
<rdar://problem/18175117> 

llvm-svn: 217083
2014-09-03 22:30:54 +00:00
Jim Ingham e7701fee0b Fix a thinko in the memory allocator that was causing us to strand an allocation range
in all but one of the AllocatedBlocks that matched the requested permissions.

Over time this would make the performance of expressions slow down considerably.

Also added a little bit of logging that was helpful in resolving the issue.

<rdar://problem/17954438>

llvm-svn: 215239
2014-08-08 20:01:41 +00:00
Jason Molenda 6076bf4edb Change ProcessGDBRemote::DoReadMemory to use the x packet to read
data if it is available.

Change ProcessGDBRemote's maximum read/write packet size from a
fixed 512 byte value to asking the remote gdb stub what its maximum
is, using up to 128kbyte sizes if that's allowed, and falling back
to 512 if the remote gdb stub doesn't advertise a max packet size.

Add a new "process plugin packet xfer-size" command that can be used
to override the maximum packet size (although not exceeding any packet
size maximum published by the remote gdb stub).
<rdar://problem/16032150> 

llvm-svn: 208058
2014-05-06 04:34:52 +00:00
Virgile Bello ffeba25652 Remove %zx in printf (only GCC supports it, not MSVC).
llvm-svn: 203349
2014-03-08 17:15:35 +00:00
Greg Clayton 15fc2be75b <rdar://problem/13892516>
LLDB can now debug across calls to exec when the architecture changes from say i386 to x86_64 (on darwin).

llvm-svn: 182345
2013-05-21 01:00:52 +00:00
Jim Ingham 55d24311f2 If the MemoryCache is asked to look up an address that is in an invalid
range, it should set the error in the Error passed into the function.

llvm-svn: 181386
2013-05-08 01:20:53 +00:00
Greg Clayton 7b0992d9cd After discussing with Chris Lattner, we require C++11, so lets get rid of the macros and just use C++11.
llvm-svn: 179805
2013-04-18 22:45:39 +00:00
Greg Clayton e01e07b6e7 Since we use C++11, we should switch over to using std::unique_ptr when C++11 is being used. To do this, we follow what we have done for shared pointers and we define a STD_UNIQUE_PTR macro that can be used and it will "do the right thing". Due to some API differences in std::unique_ptr and due to the fact that we need to be able to compile without C++11, we can't use move semantics so some code needed to change so that it can compile with either C++.
Anyone wanting to use a unique_ptr or auto_ptr should now use the "STD_UNIQUE_PTR(TYPE)" macro.

llvm-svn: 179779
2013-04-18 18:10:51 +00:00
Greg Clayton 5160ce5c72 <rdar://problem/13521159>
LLDB is crashing when logging is enabled from lldb-perf-clang. This has to do with the global destructor chain as the process and its threads are being torn down.

All logging channels now make one and only one instance that is kept in a global pointer which is never freed. This guarantees that logging can correctly continue as the process tears itself down.

llvm-svn: 178191
2013-03-27 23:08:40 +00:00
Daniel Malea d01b2953fa Resolve printf formatting warnings on Linux:
- use macros from inttypes.h for format strings instead of OS-specific types

Patch from Matt Kopec!

llvm-svn: 168945
2012-11-29 21:49:15 +00:00
Greg Clayton 43e0af06b4 Stop using the "%z" size_t modifier and cast all size_t values to uint64_t. Some platforms don't support this modification.
llvm-svn: 164148
2012-09-18 18:04:04 +00:00
Greg Clayton de0e9d04ad <rdar://problem/11193466>
Fixed an error where lldb would hang when writing memory near the end of the addres space due to an unsigned overflow.

llvm-svn: 154697
2012-04-13 20:37:20 +00:00
Greg Clayton a9f40ad80a For stepping performance I added the ability to outlaw all memory accesseses
to the __PAGEZERO segment on darwin. The dynamic loader now correctly doesn't
slide __PAGEZERO and it also registers it as an invalid region of memory. This
allows us to not make any memory requests from the local or remote debug session
for any addresses in this region. Stepping performance can improve when uninitialized
local variables that point to locations in __PAGEZERO are attempted to be read 
from memory as we won't even make the memory read or write request.

llvm-svn: 151128
2012-02-22 04:37:26 +00:00
Greg Clayton d495c5340d Added an allocated memory cache to avoid having to allocate memory over and
over when running JITed expressions. The allocated memory cache will cache 
allocate memory a page at a time for each permission combination and divvy up
the memory and hand it out in 16 byte increments. 

llvm-svn: 131453
2011-05-17 03:37:42 +00:00