讲解 Exercise 4 Understanding valgrind’s leak checker讲解 留学生Java程序

Exercise 4

Due: See website for due date.

What to submit: See website.

The theme of this exercise is automatic memory management, leak detection, and virtual memory.

1. Understanding valgrind’s leak checker

Valgrind is a tool that can aid in finding memory leaks in C programs. To that end, it per- forms an analysis similar to the “mark” phase of a traditional mark-and-sweep garbage collector right before a program exits and identifies still reachable objects and leaks. Note that at this point, the program’s main function has already returned, so any local variables defined in it have already gone out of scope.

For leaked (or lost) objects, it uses the definition prevalent for C programs:  these are objects that have been allocated but not yet freed, and there is no possible way for a legal program to access them in the future.

Read Section 4.2.8 Memory leak detection in the Valgrind Manual [URL] and then con- struct a C program leak .c that, when run with

valgrind  --leak-check=full  --show-leak-kinds=all  . /leak

produces the following output:

(the line numbers in your reconstruction need not match, but the LEAK summary should (including the number of blocks and number of bytes shown.)

2. Reverse Engineering A Memory Leak

In this part of the exercise, you will be given a post-mortem dump of a JVM’sheap that was obtained when running a program with a memory leak. The dump was produced at the point in time when the program ran out of memory because its live heap size exceeded the maximum, which can be accomplished as shown in this log:

Your task is to examine the heap dump (oom.hprof) and reverse engineer the leaky pro- gram.

To that end, you must install the Eclipse Memory Analyzer on your computer. It can be downloaded from this URL. Open the heap dump.

Requirements

•  Your program must run out of memory when run as shown above.  You should double-check  that  the  created heap  dump  matches  the provided  dump,  where “matches” is defined as follows.

•  The structure of the reachability graph of the subcomponent with the largest re- tained size should be similar in your heap dump as in the provided heap dump. (Other information such as the content of arrays may differ.)

•  You will need to write one or more classes and write code that allocates these objects and creates references between them.  You should choose the same field and class names in your program as in the heap dump, and no extra ones (we will check this). Think of field names as edge labels in the reachability graph.

•  You should investigate which classes from Java’s standard library are involved in the leak.

Hints

•  The program that was used to create the heap dump is 22 lines long (without com- ments,and including the main function), though your line numbers may differ.

•  Static inner classes are separated with a dollar sign $. For instance, A$B is the name of a static inner class called B nested in A. (Your solution should use the same class names as in the heap dump.)

•  Start with the “Leak Suspects” report, then look in Details.  Use the “List Objects ... withoutgoing references” feature to find a visualization of the objects that were part of the heap when the program ran out of memory.

•  The “dominator tree” option can also give you insight into the structure of the object graph. Zoom in on the objects that have the largest “Retained Heap” quantity.

•  Use the Java Tutor website to write small test programs and trace how the reacha- bility graph changes over time.

•  Do not forget the -Xmx64m switch when running your program, or else your pro- gram may run for several minutes before running out of memory, even if imple- mented correctly. (If implemented incorrectly, it will run forever.)

•  Do not access the  oom .hprof file through a remote file system path such as a mapped Google drive or similar. Students in the past have reported runtime errors in Eclipse MAT when trying to do that. Instead, copy it to your local computer’s file system first as a binary file. The SHA256 sum ofoom .hprof is

04df06c33e684cc8b0c4e278176ccca885d0abd71fb506e29ad25d8c331a1efa

3. Using mmap to list the entries in a ZIP file

Write a short program zipdir that displays the list of entries inside a zip file whose name is passed to the program as its first argument. For each entry it should also print its com- pression ratio as a percentage rounded to the nearest tenth of a percent. The compression ratio is defined as the ratio of the compressed size to the uncompressed size.  A sample use would be:

$   . /zipdir  heap . zip

heap1 .dot                                     41 . 9% heap1 .in                                           66 . 0% heap1 .out                                   100 . 0% heap1 .png                                      89 .4% heap2 .dot                                     36 . 8% heap2 .in                                           59 .5% heap2 .out                                   100 . 0% heap2 .png                                      92 .5% heap3 .dot                                     37 . 0% heap3 .in                                           59 .2% heap3 .out                                   100 . 0% heap3 .png                                        92 . 0%

Your program should use only the open(2), fstat(2), and mmap(2) system calls (plus any system calls needed to output the result, such as write(1) via printf.

Do not use read(2) (or higher-level functions such as fread(3), etc. that call read() internally).

The  ZIP   file   format   is   described,   among   other  places,   on   the  WikiPedia  page

https://en.wikipedia.org/wiki/ZIP_(file_format) Use the following algorithm:

• open the file with open(2) in read-only mode.

• use fstat(2) to determine the length of the file.

• use mmap(2) to map the entire file into memory in a read-only way.

•  scan from the back of the file until you find the beginning marker of the End of Central Directory Record (EOCD).

•  extract the number of central directory records in this zip archive and the start offset of the central directory.

•  Then, starting from the start offset of the central directory, examine each central directory file header and output the filename contained in it, along with the com-

pression ratio. (Hint: use the following format string for printf:

printf   ("%-25 . *s  %5 . 1f%%\n",  namelength,  name,  ratio);

•  Skip forward to the next central directory record by advancing 46 + m + n + k bytes where n is the length of the filename, m is the extra field length, and k is the file comment length contained in each central file directory header.

Simplifying assumptions/hints:

•  All multibyte integers in a ZIP file are stored in little-endian order, and — for the purposes of this exercise — you may assume that the host byte order of the machine on which your program runs is little endian as well.

•  You may use pointer arithmetic on void  * pointers, which uses a stride of 1 byte (i.e.,  it assumes that  sizeof(void)==1.   To access  16-bit or 32-bit values,  use uint16_t  * and uint32_t  *, respectively under the assumption of little endian host byte order.

•  If the given file is not a well-formed ZIP archive then the behavior. of your program can be undefined.

• Be sure to handle empty zip files that have no entries.