Add another test that locks & unlocks in FIFO order instead of stack

[pintos-anon] / TODO

-*- text -*-

Godmar says:

- In Project 2, we're missing tests that pass arguments to system calls
that span multiple pages, where some are mapped and some are not. 
An implementation that only checks the first page, rather than all pages 
that can be touched during a call to read()/write() passes all tests.

- In Project 2, we're missing a test that would fail if they assumed
that contiguous user-virtual addresses are laid out contiguously 
in memory.  The loading code should ensure that non-contiguous 
physical pages are allocated for the data segment (at least.)

- Need some tests that test that illegal accesses lead to process
termination. I have written some, will add them. In P2, obviously, 
this would require that the students break this functionality since 
the page directory is initialized for them, still it would be good 
to have.

- There does not appear to be a test that checks that they close all
fd's on exit.  Idea: add statistics & self-diagnostics code to palloc.c
and malloc.c.  Self-diagnostics code could be used for debugging.
The statistics code would report how much kernel memory is free.
Add a system call "get_kernel_memory_information".  User programs
could engage in a variety of activities and notice leaks by checking
the kernel memory statistics.

---

From: "Godmar Back" <godmar@gmail.com>
Subject: priority donation tests
To: "Ben Pfaff" <blp@cs.stanford.edu>
Date: Fri, 3 Mar 2006 11:02:08 -0500

Ben,

it seems the priority donation tests are somewhat incomplete and allow
incorrect implementations to pass with a perfect score.

We are seeing the following wrong implementations pass all tests:

- Implementations that assume that the priority of a thread waiting on
a semaphore or condition variable cannot change between when the
thread was blocked and when it is unblocked. The students implement
this by doing an insert into an ordered list on block, rather than
picking the maximum thread on unblock.

Neither of these two cases is detected; do you currently check for
these mistakes manually?

I also wrote a test case for the second scenario:
http://people.cs.vt.edu/~gback/pintos/priority-donate-sema.c
http://people.cs.vt.edu/~gback/pintos/priority-donate-sema.ck

From: "Godmar Back" <godmar@gmail.com>
Subject: multiple threads waking up at same clock tick
To: "Ben Pfaff" <blp@cs.stanford.edu>
Date: Wed, 1 Mar 2006 08:14:47 -0500

Greg Benson points out another potential TODO item for P1.

----
One thing I recall:

The alarm tests do not test to see if multiple threads are woken up if
their timers have expired.  That is, students can write a solution
that just wakes up the first thread on the sleep queue rather than
check for additional threads.  Of course, the next thread will be
woken up on the next tick.  Also, this might be hard to test.

---
Way to test this: (from Godmar Back)

Thread A with high priority spins until 'ticks' changes, then calls to
timer_sleep(X), Thread B with lower priority is then resumed, calls
set_priority to make its priority equal to that of thread A, then
calls timer_sleep(X), all of that before the next clock interrupt
arrives.

On wakeup, each thread records wake-up time and calls yield
immediately, forcing the scheduler to switch to the other
equal-priority thread. Both wake-up times must be the same (and match
the planned wake-up time.)

PS:
I actually tested it and it's hard to pass with the current ips setting.
The bounds on how quickly a thread would need to be able to return after
sleep appear too tight.  Need another idea.

From: "Godmar Back" <godmar@gmail.com>

For reasons I don't currently understand, some of our students seem
hesitant to include each thread in a second "all-threads" list and are
looking for ways to implement the advanced scheduler without one.

Currently, I believe, all tests for the mlfqs are such that all
threads are either ready or sleeping in timer_sleep(). This allows for
an incorrect implementation in which recent-cpu and priorities are
updated only for those threads that are on the alarm list or the ready
list.

The todo item would be a test where a thread is blocked on a
semaphore, lock or condition variable and have its recent_cpu decay to
zero, and check that it's scheduled right after the unlock/up/signal.

From: "Godmar Back" <godmar@gmail.com>
Subject: set_priority & donation - a TODO item
To: "Ben Pfaff" <blp@cs.stanford.edu>
Date: Mon, 20 Feb 2006 22:20:26 -0500

Ben,

it seems that there are currently no tests that check the proper
behavior of thread_set_priority() when called by a thread that is
running under priority donation.  The proper behavior, I assume, is to
temporarily drop the donation if the set priority is higher, and to
reassume the donation should the thread subsequently set its own
priority again to a level that's lower than a still active donation.

 - Godmar

From: Godmar Back <godmar@gmail.com>
Subject: project 4 question/comment regarding caching inode data
To: Ben Pfaff <blp@cs.stanford.edu>
Date: Sat, 14 Jan 2006 15:59:33 -0500

Ben,

in section 6.3.3 in the P4 FAQ, you write:

"You can store a pointer to inode data in struct inode, if you want,"

Should you point out that if they indeed do that, they likely wouldn't
be able to support more than 64 open inodes systemwide at any given
point in time.

(This seems like a rather strong limitation; do your current tests
open more than 64 files?
It would also point to an obvious way to make the projects harder by
specifically disallowing that inode data be locked in memory during
the entire time an inode is kept open.)

 - Godmar

From: Godmar Back <godmar@gmail.com>
Subject: on caching in project 4
To: Ben Pfaff <blp@cs.stanford.edu>
Date: Mon, 9 Jan 2006 20:58:01 -0500

here's an idea for future semesters.

I'm in the middle of project 4, I've started by implementing a buffer
cache and plugging it into the existing filesystem.  Along the way I
was wondering how we could test the cache.

Maybe one could adopt a similar testing strategy as in project 1 for
the MLQFS scheduler: add a function that reads "get_cache_accesses()"
and a function "get_cache_hits()".  Then create a version of pintos
that creates access traces for a to-be-determined workload.  Run an
off-line analysis that would determine how many hits a perfect cache
would have (MAX), and how much say an LRU strategy would give (MIN).
Then add a fudge factor to account for different index strategies and
test that the reported number of cache hits/accesses is within (MIN,
MAX) +/- fudge factor.

(As an aside - I am curious why you chose to use a clock-style
algorithm rather than the more straightforward LRU for your buffer
cache implementation in your sample solution. Is there a reason for
that?  I was curious to see if it made a difference, so I implemented
LRU for your cache implementation and ran the test workload of project
4 and printed cache hits/accesses.
I found that for that workload, the clock-based algorithm performs
almost identical to LRU (within about 1%, but I ran nondeterministally
with QEMU). I then reduced the cache size to 32 blocks and found again
the same performance, which raises the suspicion that the test
workload might not force any cache replacement, so the eviction
strategy doesn't matter.)

Godmar Back <godmar@gmail.com> writes:

> in your sample solution to P4, dir_reopen does not take any locks when
> changing a directory's open_cnt. This looks like a race condition to
> me, considering that dir_reopen is called from execute_process without
> any filesystem locks held.

* Get rid of rox--causes more trouble than it's worth

* Reconsider command line arg style--confuses everyone.

* Finish writing tour.

via Godmar Back:

* Get rid of mmap syscall, add sbrk.

* page-linear, page-shuffle VM tests do not use enough memory to force
  eviction.  Should increase memory consumption.

* Add FS persistence test(s).

* process_death test needs improvement

* Internal tests.

* Improve automatic interpretation of exception messages.

* Userprog project:

  - Mark read-only pages as actually read-only in the page table.  Or,
    since this was consistently rated as the easiest project by the
    students, require them to do it.

  - Don't provide per-process pagedir implementation but only
    single-process implementation and require students to implement
    the separation?  This project was rated as the easiest after all.
    Alternately we could just remove the synchronization on pid
    selection and check that students fix it.

* Filesys project:

  - Need a better way to measure performance improvement of buffer
    cache.  Some students reported that their system was slower with
    cache--likely, Bochs doesn't simulate a disk with a realistic
    speed.

* Documentation:

  - Add "Digging Deeper" sections that describe the nitty-gritty x86
    details for the benefit of those interested.

  - Add explanations of what "real" OSes do to give students some
    perspective.

* Assignments:

  - Add extra credit:

    . Low-level x86 stuff, like paged page tables.

    . Specifics on how to implement sbrk, malloc.

    . Other good ideas.

    . opendir/readdir/closedir

    . everything needed for getcwd()

To add partition support:

- Find four partition types that are more or less unused and choose to
  use them for Pintos.  (This is implemented.)

- Bootloader reads partition tables of all BIOS devices to find the
  first that has the "Pintos kernel" partition type.  (This is
  implemented.)  Ideally the bootloader would make sure there is
  exactly one such partition, but I didn't implement that yet.

- Bootloader reads kernel into memory at 1 MB using BIOS calls.  (This
  is implemented.)

- Kernel arguments have to go into a separate sector because the
  bootloader is otherwise too big to fit now?  (I don't recall if I
  did anything about this.)

- Kernel at boot also scans partition tables of all the disks it can
  find to find the ones with the four Pintos partition types (perhaps
  not all exist).  After that, it makes them available to the rest of
  the kernel (and doesn't allow access to other devices, for safety).

- "pintos" and "pintos-mkdisk" need to write a partition table to the
  disks that they create.  "pintos-mkdisk" will need to take a new
  parameter specifying the type.  (I might have partially implemented
  this, don't remember.)

- "pintos" should insist on finding a partition header on disks handed
  to it, for safety.

- Need some way for "pintos" to assemble multiple disks or partitions
  into a single image that can be copied directly to a USB block
  device.  (I don't know whether I came up with a good solution yet or
  not, or whether I implemented any of it.)

To add USB support:

- Needs to be able to scan PCI bus for UHCI controller.  (I
  implemented this partially.)

- May want to be able to initialize USB controllers over CardBus
  bridges.  I don't know whether this requires additional work or if
  it's useful enough to warrant extra work.  (It's of special interest
  for me because I have a laptop that only has USB via CardBus.)

- There are many protocol layers involved: SCSI over USB-Mass Storage
  over USB over UHCI over PCI.  (I may be forgetting one.)  I don't
  know yet whether it's best to separate the layers or to merge (some
  of) them.  I think that a simple and clean organization should be a
  priority.

- VMware can likely be used for testing because it can expose host USB
  devices as guest USB devices.  This is safer and more convenient
  than using real hardware for testing.

- Should test with a variety of USB keychain devices because there
  seems to be wide variation among them, especially in the SCSI
  protocols they support.  Should try to use a "lowest-common
  denominator" SCSI protocol if any such thing really exists.

- Might want to add a feature whereby kernel arguments can be given
  interactively, rather than passed on-disk.  Needs some though.