Notes

memory leaks
race conidtions
security
exec on non-existent or non-elf, return -1
not re-claiming user heap (should handle for it)
16 concurrent processes
1 MB
only 10 seconds to run the test
HEAP
fragments the heap under any circumstance
we can take advantage of undefined behavior


-----
copy and write fork
read only FS (implement write)
i-nodes
implement the login process (security authentication)
scheduler
networking
security
shells
heap





____



you try to reference kernel address kills the thread -1
you try to read, and give kernel to buffer in the address space, return -1 for read system call
exit from thread any kind of time it tries to do shitty thing

copy user code
UD2
????????
we own user directory


Reading from Pipe should not return 0 unless it blocks forever?
Make sure you can't write or read in kernel memory, seek, join
READ
- return -1

WRITE
- security

SEEKING
- past end of file, return -1
- end of file, read will reutrn 0
- cannot seek on pipe

JOIN
- only join by parent
- can't join more than once
- if parent exit, than no joins can occur
- join itself returns -1
- join itself deadlock


do anything to invalid file descriptor, return -1
you are only allowed to have 100 file descriptors
we don't have processes anymore???
semaphore table only 2015, ever, can't be reusing
fd is only 100 concurrent (smallest available)
2000 threads ever (global limit), physical memory for

FORK
- creates a new thread with a new address space + copy + fd + semaphore table
THREAD
-

we don't have to handle multi threads forking???
EXEC
- create a new address space, share fd and semaphore table (THIS IS DIFFERENT THAN FORK)

Some stuff with ELF file and Magic number

PIPE, something with close to kernel return -1

THREAD STACK
- point to a function that leaves the thread