xv6-65oo2/labs/cow.html
Frans Kaashoek b02ef59e14 x
2019-08-01 16:52:38 -04:00

110 lines
3.6 KiB
HTML

<html>
<head>
<title>Lab: Copy-on-Write Fork for xv6</title>
<link rel="stylesheet" href="homework.css" type="text/css" />
</head>
<body>
<h1>Lab: Copy-on-Write Fork for xv6</h2>
<p>
Your task is implement copy-on-write fork in the xv6 kernel. You are
done if your modified kernel executes both the cow and usertests
programs successfully.
<h2>The problem</h2>
The fork() system call in xv6 copies all of the parent process's
user-space memory into the child. If the parent is large, copying can
take a long time. In addition, the copies often waste memory; in many
cases neither the parent nor the child modifies a page, so that in
principle they could share the same physical memory. The inefficiency
is particularly clear if the child calls exec(), since then most of
the copied pages are thrown away without ever being used. Of course,
sometimes both child and parent modify memory at the same virtual
address after a fork(), so for some pages the copying is truly needed.
<h2>The solution</h2>
The goal of copy-on-write (COW) fork() is to defer allocating and
copying physical memory pages for the child until they are actually
needed, in the hope that they may never be needed.
<p>
COW fork() creates just a pagetable for the child, with PTEs for user
memory pointing to the parent's physical pages. COW fork() marks all
the user PTEs in both parent and child as read-only. When either
process tries to write one of these COW pages, the CPU will force a
page fault. The kernel page-fault handler detects this case, allocates
a page of physical memory for the faulting process, copies the
original page into the new page, and modifies the relevant PTE in the
faulting process to refer to the new page, this time with the PTE
marked writeable. When the page fault handler returns, the user
process will be able to write its copy of the page.
<p>
COW fork() makes freeing of the physical pages that implement user
memory a little trickier. A given physical page may be referred to by
multiple processes' page tables, and should be freed when the last
reference disappears.
<h2>The cow test program</h2>
To help you test your implementation, we've provided an xv6 program
called cow (source in user/cow.c). cow runs various tests, but
even the first will fail on unmodified xv6. Thus, initially, you
will see:
<pre>
$ cow
simple: fork() failed
$
</pre>
The "simple" test allocates more than half of available physical
memory, and then fork()s. The fork fails because there is not enough
free physical memory to give the child a complete copy of the parent.
<p>
When you are done, your kernel should be able to run both cow and
usertests. That is:
<pre>
$ cow
simple: ok
simple: ok
three: zombie!
ok
three: zombie!
ok
three: zombie!
ok
file: ok
ALL COW TESTS PASSED
$ usertests
...
ALL TESTS PASSED
$
</pre>
<h2>Hints</h2>
Here's one reasonable plan of attack. Modify uvmcopy() to map the
parent's physical pages into the child, instead of allocating new
pages, and clear PTE_W in the PTEs of both child and parent.
Modify usertrap() to recognize a page fault. When a page fault occurs
on a COW page, allocate a new page with kalloc(), copy the old page to
the new page, and install the new page in the PTE with PTE_W set.
Next, ensure that each physical page is freed when the last PTE
reference to it goes away (but not before!), perhaps by implementing
reference counts in kalloc.c. Finally, modify copyout() to use the
same scheme as page faults when it encounters a COW page.
<p>
It may be useful to have a way to record, for each PTE, whether it is
a COW mapping. You can use the RSW (reserved for software) bits in
the RISC-V PTE for this.
</body>
</html>