CS 551 Systems Programming

CS 551 Systems Programming, Fall 2024 Programming Project 1

Out: 10/13/2024 Sun.

Due: 10/26/2024 Sat. 23:59:59 In this project your are going to implement a custom memory manager that manages heapmemory allocation at program level. Here are the reasons why we need a custom memory managerin C.

  • The C memory allocation functions malloc and free bring performance overhead: thesecalls may lead to a switch between user-space and kernel-space. For programs that do frequent memory allocation/deallocation, using malloc/free will degrade the performance.
  • A custom memory manager allows for detection of memory misuses, such as memory overflow, memory leak and double deallocating a pointer.The goal of this project is to allow students to practice on C programming, especially on bitwiseoperations, pointer operation, linked list, and memory management.User programYour memory manager

(a library)OS kernelUser programOS kernelmalloc/freemalloc/free (interposed version)malloc/free

Figure 1: Overview

1 Overview

Figure 1 depicts where the memory manager locates in a system. The memory manager basicallypreallocates chunks of memory, and performs management on them based on the user programmemory allocation/deallocation requests.

We use an example to illustrate how your memory manager is supposed to work. Supposememory allocations in the memory manager are 16 bytes aligned1 .

  • After the user program starts, the first memory allocation from the program requests 12 bytesof memory (malloc(12)). Prior to this first request, your memory manager is initialized1In other words, the memory manager always allocates memory that is a multiple of 16 bytes. In the base code,this is controlled by the MEM ALIGNMENT BOUNDARY macro defined in memory manager.hwith a batch of memory slots2 , each of which has a size of 16 bytes. Memory manager returnsthe first slot of the batch to the user program.
  • Then the user program makes a second request (malloc(10)). Because the memory allocation is 16 bytes aligned, the memory manager should return a chunk of memory of 16bytes. This time, because there are still available 16-byte-slots in the allocated batch, thememory manager simply return the second slot in the allocated batch to fulfill the requestwithout interacting with the kernel.
  • The user program makes 6 subsequent memory requests, all of which are for memory lessthan 16 bytes. The memory manager simply returns each of the rest 6 free 16-byte-slots tofulfill the requests. And for the implementation of this project, assume youwill only getrequests for less than or equal to 16 bytes memory.
  • The user program makes the 9th request (malloc(7)). Because there is no free slots avail

able, the manager allocates another batch of memory slots of 16 bytes, and returns the first

slot in this batch to the user program.

  • Suppose the 10th memory request from the user program is malloc(28). The managershould return a memory chunk of 32 bytes (remember memory allocation is 16 bytes aligned).Because there is no memory list of 32-byte-slots, the manager has to allocate a batch ofmemory slots of 32 bytes, and returns the first slot to fulfill this request. At this moment,

the memory list of 32-byte-slots has only one memory batch.

  • The memory manager organizes memory batches that have the same slot size using linkedlist, and the resulting list is called a memory batch list, or a memory list.
  • Memory lists are also linked together using linked list. So the default memory list with slotsize of 16 bytes is linked with the newly created 32 bytes slot size memory list.
  • The manager uses a bitmap to track and manage slots allocation/deallocation for eachmemory batch list.

t is easy to see that with such a mechanism, the memory manager not only improves programperformance by reducing the number of kernel/user space switches, utalso tracks all the memoryallocation/deallocation so that it can detect memory misuse such as double freeing. The memory

nager can also add guard bytes at the end of each memory slot to detect memory overflow (just

an example, adding guard bytes is not required by this project.)

2 What to do

2.1 Download the base code Download the base code from Assignments section in the Brightspace. You will need to add yourimplementation into this base code.2In the base code, the number of memory slots in a batch is controlled by the MEM BATCH SLOT COUNT macrodefined in memory manager.h2.2 Complete the bitmap operation functions (25 points)

Complete the implementation of the functions in bitmap.c.

  • int bitmap find first bit(unsigned char * bitmap, int size, int val)This function finds the position (starting from 0) of the first bit whose value is “val” in the

“bitmap”. The val could be either 0 or 1.

Suppose

unsigned char bitmap[] = {0xF7, 0xFF};

Then a call as following

bitmap_find_first_bit(bitmap, sizeof(bitmap), 0);

finds the first bit whose value is 0. The return value should be 3 in this case.

  • int bitmap set bit(unsigned char * bitmap, int size, int target pos)

This function sets the “target pos”-th bit (starting from 0) in the “bitmap” to 1.

Suppose

unsigned char bitmap[] = {0xF7, 0xFF};

Then a call as following

bitmap_set_bit(bitmap, sizeof(bitmap), 3);

sets bit-3 to 1. After the call the content of the bitmap is {0xFF, 0xFF};

  • int bitmap clear bit(unsigned char * bitmap, int size, int target pos)

This function sets the “target pos”-th bit (starting from 0) in the “bitmap” to 0.

  • int bitmap bit is set(unsigned char * bitmap, int size, int pos)

This function tests if the “pos”-th bit (starting from 0) in the “bitmap” is 1.

Suppose

unsigned char bitmap[] = {0xF7, 0xFF};

Then a call as following

bitmap_bit_is_set(bitmap, sizeof(bitmap), 3);

returns 0, because the bit-3 in the bitmap is 0.

  • int bitmap print bitmap(unsigned char * bitmap, int size)

This function prints the content of a bitmap in starting from the first bit, and insert a space

every 4 bits.

Suppose

unsigned char bitmap[] = {0xA7, 0xB5};

Then a call to the function would print the content of the bitmap as

1110 0101 1010 1101

The implementation of this function is given.2.3 Complete the memory manager implementation (70 points) In memory manager.h two important structures are defined:

  • structstru mem batch: structure definition of the a memory batch (of memory slots).
  • structstru mem list: structure definition of a memory list (of memory batches).To better assist you understand how a memory manager is supposed to organize the memoryusing the two data structures, Figure 2 shows an example of a possible snapshot of the memorymanager’s data structures.Figure 2: An example of data structuresBasically there are two kinds of linked list:

A list of memory batches (with a certain slot size): as shown in the previous example, this

ist expands when there is no free slot available. The memory manager adds a new batch atthe end of the list.

  • A list of memory batch list: this list expands when a new slot size comes in.You will need to implement the functions in memory manager.c:
  • void * mem mngr alloc(size t size)– This is the memory allocation function of the memory manager. It is used in the same

way as malloc().

– Provide your implementation.

– The macro MEM ALIGNMENT BOUNDARY (defined in memory manager.h) controlshow memory allocation is aligned. For this project, we use 16 byte aligned.Butyour implementation should be able to handle any alignment. One reason to define the alignment as a macro is to allow for easy configuration. When grading wewill test alignment that is multiples of 16 by changing the definition of the macro

MEM ALIGNMENT BOUNDARY to 8, 32, ....– The macro MEM BATCH SLOT COUNT (defined in memory manager.h) controls thenumber of slots in a memory batch. For this project this number is set to 8. But your代 写CS 551 Systems Programming implementation should also work if we change to the macro MEM BATCH SLOT COUNT

to other values. When grading, we will test the cases when MEM BATCH SLOT COUNTis set to multiples of 8 (i.e., 8, 16, 24, ...).

– When there are multiple free slots, return the slot using the first-fit policy(i.e, the one

with smallest address).

– Remember to clear the corresponding bit in free slots bitmap after a slot is allo

cated to user program.

– Do not use array for bitmap, because you never now how many bits you will need. Use

– Remember to expand the bitmap when a new batch is added to a list, and keep the oldcontent after expansion.

– Create a new memory list to handle the request if none of the existing memory list candeal with the requested size.

– Add this memory list into the list of memory lists.

– This function should support allocation size up to 5 times the alignment size, e.g., 80bytes for 16 byte alignment.

  • void mem mngr free(void * ptr)

– This is the memory free function of the memory manager. It is used in the same wayas free().

– Provide your implementation.

– The ptr should be a starting address of an assigned slot, report(print out) error if itis not (three possible cases:

1: ptr is the starting address of an unassigned slot - double freeing;

2: ptr is outside of memory managed by the manager;

3: ptr is not the starting address of any slot).

– Remeber to set the corresponding bit in free slots bitmap after a slot is freed, so

that it can be used again.

– Search through all the memory lists to find out which memory batch the ptr associatedslot belongs to.• void mem mngr init(void)

– This function is called by user program when it starts.

– Initialize the lists of memory batch list with one default batch list. The slot size of thisdefault batch list is 16 bytes.

– Initialize this default list with one batch of memory slots.

– Initialize the bitmap of this default list with all bits set to 1, which suggests that allslots in the batch are free to be allocated.

  • void mem mngr leave(void)

– This function is called by user program when it quits. It basically frees all the heap

memory allocated.

– Provide your implementation.

– Don’t forget to free all the memory lists to avoid memory leak.

  • void mem mngr print snapshot(void)

This function has already been implemented. It prints out the current status of the memorymanager. Reading this function may help you understand how the memory manager organizes the memory. Do not change the implementation of this function. It will be used tohelp the grading.

2.4 Writing a makefile (5 points)

Write a makefile to generate

  • memory manager.o
  • bitmap.o
  • a static library memory manager.a, which contains the previous relocatable object files.This library will be linked to our test program for testing.

2.5 Log and submit your work

Log your work: besides the files needed to build your project, you must also include a READMEfile which minimally contains your name and B-number. Additionally, it can contain the following:

  • The status of your program (especially, if not fully complete).
  • Bonus is implemented or not.
  • A description of how your code works, if that is not completely clear by reading the code(note that this should not be necessary, ideally your code should be self-documenting).
  • Possibly a log of test cases which work and which don’t work.
  • Any other material you believe is relevant to the grading of your project.Compress the files: compress the following into a ZIP file:
  • bitmap.c
  • common.h
  • interposition.h
  • memory manager.c
  • memory manager.h
  • Makefile
  • README

Name the ZIP file based on your BU email ID. For example, if your BU email is “abc@binghamton.edu”,then the zip file should be “proj1 abc.zip”.

Submission: submit the ZIP file to Brightspace before the deadline.

2.6 Grading guidelines

(1) Prepare the ZIP file on a Linux machine. If your zip file cannot be uncompressed, 5 pointsoff.

(2) If the submitted ZIP file/source code files included in the ZIP file are not named as specifiedabove (so that it causes problems for TA’s automated grading scripts), 10 points off.(3) If the submitted code does not compile:

1 TA will try to fix the problem (for no more than 3 minutes);

2 if (problem solved)3 1%-10% points off (based on how complex the fix is, TA’s discretion);4 else

5 TA may contact the student by email or schedule a demo to fix the problem;6 if (problem solved)

7 11%-20% points off (based on how complex the fix is, TA’s discretion);8 else

9 All points off;So in the case that TA contacts you to fix a problem, please respond to TA’s email promptlyor show up at the demo appointment on time; otherwise the line 9 above will be effective.

(4) If the code is not working as required in this spec, the TA should take points based on theassigned full points of the task and the actual problem.

(5) Late penalty: Day1 10%, Day2 20%, Day3 40%, Day4 60%, Day5 80%(6) Lastly but not the least, stick to the collaboration policy stated in the syllabus: you maydiscuss with your fellow students, but code should absolutely be kept private. Any violationto the Academic Honesty Policy and University Academic Policy, will result 0 for the work.

posted @ 2024-10-19 12:45  hellyou  阅读(9)  评论(0)  编辑  收藏  举报