static int allocate_ports(struct mgcp_endpoint *endp)
{
if (allocate_port(endp, &endp->net_end, &endp->cfg->net_ports,
mgcp_bind_net_rtp_port) != 0)
return -1;
if (allocate_port(endp, &endp->bts_end, &endp->cfg->bts_ports,
mgcp_bind_bts_rtp_port) != 0) {
mgcp_rtp_end_reset(&endp->net_end);
return -1;
}
if (endp->cfg->transcoder_ip && endp->tcfg->trunk_type == MGCP_TRUNK_VIRTUAL) {
if (allocate_port(endp, &endp->trans_net,
&endp->cfg->transcoder_ports,
mgcp_bind_trans_net_rtp_port) != 0) {
mgcp_rtp_end_reset(&endp->net_end);
mgcp_rtp_end_reset(&endp->bts_end);
return -1;
}
if (allocate_port(endp, &endp->trans_bts,
&endp->cfg->transcoder_ports,
mgcp_bind_trans_bts_rtp_port) != 0) {
mgcp_rtp_end_reset(&endp->net_end);
mgcp_rtp_end_reset(&endp->bts_end);
mgcp_rtp_end_reset(&endp->trans_net);
return -1;
}
/* remember that we have set up transcoding */
endp->is_transcoded = 1;
}
return 0;
}
void
add_outbound_port( list_element **switches, uint64_t dpid, uint16_t port_no ) {
assert( port_no != 0 );
info( "Adding a port: dpid = %#" PRIx64 ", port = %u", dpid, port_no );
// lookup switch
switch_info *sw = lookup_switch( *switches, dpid );
if ( sw == NULL ) {
sw = allocate_switch( dpid );
append_to_tail( switches, sw );
}
port_info *new_port = allocate_port( dpid, port_no );
append_to_tail( &sw->ports, new_port );
}
extern void
system_call_handler(void)
{
register int schedule=0;
/*!< System calls may set this variable to 1. The variable is used as
input to the scheduler to indicate that scheduling is not necessary. */
/* Reset the interrupt flag indicating that the context of the caller was
saved by the system call routine. */
thread_table[cpu_private_data.thread_index].data.registers.from_interrupt=0;
switch(SYSCALL_ARGUMENTS.rax)
{
case SYSCALL_PAUSE:
{
register int tmp_thread_index;
unsigned long timer_ticks=SYSCALL_ARGUMENTS.rdi;
/* Set the return value before doing anything else. We will switch to a new
thread very soon! */
SYSCALL_ARGUMENTS.rax=ALL_OK;
if (0 == timer_ticks)
{
/* We should not wait if we are asked to wait for less then one tick. */
break;
}
/* Get the current thread. */
tmp_thread_index=cpu_private_data.thread_index;
/* Force a re-schedule. */
schedule=1;
/* And insert the thread into the timer queue. */
/* The timer queue is a linked list of threads. The head (first entry)
(thread) in the list has a list_data field that holds the number of
ticks to wait before the thread is made ready. The next entries (threads)
has a list_data field that holds the number of ticks to wait after the
previous thread is made ready. This is called to use a delta-time and
makes the code to test if threads should be made ready very quick. It
also, unfortunately, makes the code that insert code into the queue
rather complex. */
/* If the queue is empty put the thread as only entry. */
if (-1 == timer_queue_head)
{
thread_table[tmp_thread_index].data.next=-1;
thread_table[tmp_thread_index].data.list_data=timer_ticks;
timer_queue_head=tmp_thread_index;
}
else
{
/* Check if the thread should be made ready before the head of the
previous timer queue. */
register int curr_timer_queue_entry=timer_queue_head;
if (thread_table[curr_timer_queue_entry].data.list_data>timer_ticks)
{
/* If so set it up as the head in the new timer queue. */
thread_table[curr_timer_queue_entry].data.list_data-=timer_ticks;
thread_table[tmp_thread_index].data.next=curr_timer_queue_entry;
thread_table[tmp_thread_index].data.list_data=timer_ticks;
timer_queue_head=tmp_thread_index;
}
else
{
register int prev_timer_queue_entry = curr_timer_queue_entry;
/* Search until the end of the queue or until we found the right spot. */
while((-1 != thread_table[curr_timer_queue_entry].data.next) &&
(timer_ticks>=thread_table[curr_timer_queue_entry].data.list_data))
{
timer_ticks-=thread_table[curr_timer_queue_entry].data.list_data;
prev_timer_queue_entry=curr_timer_queue_entry;
curr_timer_queue_entry=thread_table[curr_timer_queue_entry].data.next;
}
if (timer_ticks>=thread_table[curr_timer_queue_entry].data.list_data)
{
/* Insert the thread into the queue after the existing entry. */
thread_table[tmp_thread_index].data.next=
thread_table[curr_timer_queue_entry].data.next;
thread_table[curr_timer_queue_entry].data.next=tmp_thread_index;
thread_table[tmp_thread_index].data.list_data=timer_ticks-
thread_table[curr_timer_queue_entry].data.list_data;
}
else
{
/* Insert the thread into the queue before the existing entry. */
thread_table[tmp_thread_index].data.next=
curr_timer_queue_entry;
thread_table[prev_timer_queue_entry].data.next=tmp_thread_index;
thread_table[tmp_thread_index].data.list_data=timer_ticks;
thread_table[curr_timer_queue_entry].data.list_data-=timer_ticks;
}
}
}
break;
}
case SYSCALL_TIME:
{
/* Returns the current system time to the program. */
SYSCALL_ARGUMENTS.rax=system_time;
break;
}
case SYSCALL_FREE:
{
SYSCALL_ARGUMENTS.rax=kfree(SYSCALL_ARGUMENTS.rdi);
break;
}
case SYSCALL_ALLOCATE:
{
/* Check the flags. */
if (0!=SYSCALL_ARGUMENTS.rsi & ~(ALLOCATE_FLAG_READONLY|ALLOCATE_FLAG_EX))
{
/* Return if the flags were not properly set. */
SYSCALL_ARGUMENTS.rax = ERROR;
break;
}
SYSCALL_ARGUMENTS.rax=kalloc(
SYSCALL_ARGUMENTS.rdi,
thread_table[cpu_private_data.thread_index].data.owner,
SYSCALL_ARGUMENTS.rsi & (ALLOCATE_FLAG_READONLY|ALLOCATE_FLAG_EX));
break;
}
case SYSCALL_ALLOCATEPORT:
{
int port=allocate_port(SYSCALL_ARGUMENTS.rdi,
thread_table[cpu_private_data.thread_index].
data.owner);
/* Return an error if a port cannot be allocated. */
if (port<0)
{
SYSCALL_ARGUMENTS.rax=ERROR;
break;
}
SYSCALL_ARGUMENTS.rax=port;
break;
}
case SYSCALL_FINDPORT:
{
int port;
unsigned long process=SYSCALL_ARGUMENTS.rsi;
/* Return an error if the process argument is wrong. */
if (process >= MAX_NUMBER_OF_PROCESSES)
{
SYSCALL_ARGUMENTS.rax=ERROR;
break;
}
port=find_port(SYSCALL_ARGUMENTS.rdi,
process);
if (port<0)
{
SYSCALL_ARGUMENTS.rax=ERROR;
break;
}
SYSCALL_ARGUMENTS.rax=port;
break;
}
case SYSCALL_GETPID:
{
SYSCALL_ARGUMENTS.rax =
thread_table[cpu_private_data.thread_index].data.owner;
break;
}
case SYSCALL_GETSCANCODE:
{
/* Check if there is data in the scan code buffer. */
if (keyboard_scancode_high_marker!=keyboard_scancode_low_marker)
{
/* There is data in the buffer. Get it! */
SYSCALL_ARGUMENTS.rax=keyboard_scancode_buffer
[(keyboard_scancode_low_marker++)&(KEYBOARD_BUFFER_SIZE-1)];
}
else
{
/* No data in the buffer. We will block waiting for data. */
register int current_thread_index;
/* Set the default return value to be an error. */
SYSCALL_ARGUMENTS.rax=ERROR;
/* Take the current thread out of the ready queue. */
current_thread_index=cpu_private_data.thread_index;
/* Tell the scheduler that it will have to reschedule. */
schedule=1;
thread_queue_enqueue(&keyboard_blocked_threads, current_thread_index);
}
break;
}
#include "syscall.c"
default:
{
/* No system call defined. */
SYSCALL_ARGUMENTS.rax=ERROR_ILLEGAL_SYSCALL;
}
}
#include "scheduler.c"
}
void
initialize(void)
{
register int i;
/* Loop over all threads in the thread table and reset the owner. */
for(i=0; i<MAX_NUMBER_OF_THREADS; i++)
{
thread_table[i].data.owner=-1; /* -1 is an illegal process_table index.
We use that to show that the thread
is dormant. */
}
/* Loop over all processes in the thread table and mark them as not
executing. */
for(i=0; i<MAX_NUMBER_OF_PROCESSES; i++)
{
process_table[i].threads=0; /* No executing process has less than 1
thread. */
}
/* Initialize the ready queue. */
thread_queue_init(&ready_queue);
/* Initialize the list of blocked threads waiting for the keyboard. */
thread_queue_init(&keyboard_blocked_threads);
/* Calculate the number of pages. */
memory_pages = memory_size/(4*1024);
{
/* Calculate the number of frames occupied by the kernel and executable
images. */
const register int k=first_available_memory_byte/(4*1024);
/* Mark the pages that are used by the kernel or executable images as taken
by the kernel (-2 in the owner field). */
for(i=0; i<k; i++)
{
page_frame_table[i].owner=-2;
page_frame_table[i].free_is_allowed=0;
}
/* Loop over all the rest page frames and mark them as free (-1 in owner
field). */
for(i=k; i<memory_pages; i++)
{
page_frame_table[i].owner=-1;
page_frame_table[i].free_is_allowed=1;
}
/* Mark any unusable pages as taken by the kernel. */
for(i=memory_pages; i<MAX_NUMBER_OF_FRAMES; i++)
{
page_frame_table[i].owner=-2;
page_frame_table[i].free_is_allowed=0;
}
}
/* Go through the linked list of executable images and verify that they
are correct. At the same time build the executable_table. */
{
const struct executable_image* image;
for (image=ELF_images_start; 0!=image; image=image->next)
{
unsigned long image_size;
/* First calculate the size of the image. */
if (0 != image->next)
{
image_size = ((char *) (image->next)) - ((char *) image) -1;
}
else
{
image_size = ((char *) ELF_images_end) - ((char *) image) - 1;
}
/* Check that the image is an ELF image and that it is of the
right type. */
if (
/* EI_MAG0 - EI_MAG3 have to be 0x7f 'E' 'L' 'F'. */
(image->elf_image.e_ident[EI_MAG0] != 0x7f) ||
(image->elf_image.e_ident[EI_MAG1] != 'E') ||
(image->elf_image.e_ident[EI_MAG2] != 'L') ||
(image->elf_image.e_ident[EI_MAG3] != 'F') ||
/* Check that the image is a 64-bit image. */
(image->elf_image.e_ident[EI_CLASS] != 2) ||
/* Check that the image is a little endian image. */
(image->elf_image.e_ident[EI_DATA] != 1) ||
/* And that the version of the image format is correct. */
(image->elf_image.e_ident[EI_VERSION] != 1) ||
/* NB: We do not check the ABI or ABI version. We really should
but currently those fields are not set properly by the build
tools. They are both set to zero which means: System V ABI,
third edition. However, the ABI used is clearly not System V :-) */
/* Check that the image is executable. */
(image->elf_image.e_type != 2) ||
/* Check that the image is executable on AMD64. */
(image->elf_image.e_machine != 0x3e) ||
/* Check that the object format is corrent. */
(image->elf_image.e_version != 1) ||
/* Check that the processor dependent flags are all reset. */
(image->elf_image.e_flags != 0) ||
/* Check that the length of t he header is what we expect. */
(image->elf_image.e_ehsize != sizeof(struct Elf64_Ehdr)) ||
/* Check that the size of the program header table entry is what
we expect. */
(image->elf_image.e_phentsize != sizeof(struct Elf64_Phdr)) ||
/* Check that the number of entries is reasonable. */
(image->elf_image.e_phnum < 0) ||
(image->elf_image.e_phnum > 8) ||
/* Check that the entry point is within the image. */
(image->elf_image.e_entry < 0) ||
(image->elf_image.e_entry >= image_size) ||
/* Finally, check that the program header table is within the image. */
(image->elf_image.e_phoff > image_size) ||
((image->elf_image.e_phoff +
image->elf_image.e_phnum * sizeof(struct Elf64_Phdr)) > image_size )
)
{
/* There is something wrong with the image. */
while (1)
{
kprints("Kernel panic! Corrupt executable image.\n");
}
continue;
}
/* Now check the program header table. */
{
int program_header_index;
struct Elf64_Phdr* program_header = ((struct Elf64_Phdr*)
(((char*) &(image->elf_image)) +
image->elf_image.e_phoff));
unsigned long memory_footprint_size = 0;
for (program_header_index = 0;
program_header_index < image->elf_image.e_phnum;
program_header_index++)
{
/* First sanity check the entry. */
if (
/* Check that the segment is a type we can handle. */
(program_header[program_header_index].p_type < 0) ||
(!((program_header[program_header_index].p_type == PT_NULL) ||
(program_header[program_header_index].p_type == PT_LOAD) ||
(program_header[program_header_index].p_type == PT_PHDR))) ||
/* Look more carefully into loadable segments. */
((program_header[program_header_index].p_type == PT_LOAD) &&
/* Check if any flags that we can not handle is set. */
(((program_header[program_header_index].p_flags & ~7) != 0) ||
/* Check if sizes and offsets look sane. */
(program_header[program_header_index].p_offset < 0) ||
(program_header[program_header_index].p_vaddr < 0) ||
(program_header[program_header_index].p_filesz < 0) ||
(program_header[program_header_index].p_memsz < 0) ||
/* Check if the segment has an odd size. We require the
segment size to be an even multiple of 8. */
(0 != (program_header[program_header_index].p_memsz&7)) ||
(0 != (program_header[program_header_index].p_filesz&7)) ||
/* Check if the segment goes beyond the image. */
((program_header[program_header_index].p_offset +
program_header[program_header_index].p_filesz) > image_size)))
)
{
while (1)
{
kprints("Kernel panic! Corrupt segment.\n");
}
}
/* Check that all PT_LOAD segments are contiguous starting from
address 0. Also, calculate the memory footprint of the image. */
if (program_header[program_header_index].p_type == PT_LOAD)
{
if (program_header[program_header_index].p_vaddr !=
memory_footprint_size)
{
while (1)
{
kprints("Kernel panic! Executable image has illegal memory layout.\n");
}
}
memory_footprint_size += program_header[program_header_index].p_memsz;
}
}
executable_table[executable_table_size].memory_footprint_size =
memory_footprint_size;
}
executable_table[executable_table_size].elf_image = &(image->elf_image);
executable_table_size += 1;
kprints("Found an executable image.\n");
if (executable_table_size >= MAX_NUMBER_OF_PROCESSES)
{
while (1)
{
kprints("Kernel panic! Too many executable images found.\n");
}
}
}
}
/* Check that actually some executable files are found. Also check that the
thread structure is of the right size. The assembly code will break if it
is not. Finally, initialize memory protection. You will implement memory
protection in task A4. */
if ((0 >= executable_table_size) ||
(1024 != sizeof(union thread)))
{
while (1)
{
kprints("Kernel panic! Can not boot.\n");
}
}
initialize_memory_protection();
initialize_ports();
initialize_thread_synchronization();
/* All sub-systems are now initialized. Kernel areas can now get the right
memory protection. */
{
/* Use the kernel's ELF header. */
struct Elf32_Phdr* program_header = ((struct Elf32_Phdr*)
(((char*) (0x00100000)) +
((struct Elf32_Ehdr*)0x00100000)->
e_phoff));
/* Traverse the program header. */
short number_of_program_header_entries =
((struct Elf32_Ehdr*)0x00100000)->e_phnum;
int i;
for(i=0; i<number_of_program_header_entries; i++)
{
if (PT_LOAD == program_header[i].p_type)
{
/* Set protection on each segment. */
update_memory_protection(kernel_page_table_root,
program_header[i].p_vaddr,
program_header[i].p_memsz,
(program_header[i].p_flags&7) | PF_KERNEL);
}
}
}
/* Start running the first program in the executable table. */
/* Use the ELF program header table and copy the right portions of the
image to memory. This is done by prepare_process. */
{
struct prepare_process_return_value prepare_process_ret_val =
prepare_process(executable_table[0].elf_image,
0,
executable_table[0].memory_footprint_size);
if (0 == prepare_process_ret_val.first_instruction_address)
{
while (1)
{
kprints("Kernel panic! Can not start process 0!\n");
}
}
/* Start executable program 0 as process 0. At this point, there are no
processes so we can just grab entry 0 and use it. */
process_table[0].parent=-1; /* We put -1 to indicate that there is no
parent process. */
process_table[0].threads=1;
/* all processes should start with an allocated port with id zero */
if (-1 == allocate_port(0,0))
{
while(1)
{
kprints("Kernel panic! Can not initialize the IPC system!\n");
}
}
/* Set the page table address. */
process_table[0].page_table_root =
prepare_process_ret_val.page_table_address;
cpu_private_data.page_table_root =
prepare_process_ret_val.page_table_address;
/* We need a thread. We just take the first one as no threads are running or
have been allocated at this point. */
thread_table[0].data.owner=0; /* 0 is the index of the first process. */
/* We reset all flags and enable interrupts */
thread_table[0].data.registers.integer_registers.rflags=0x200;
/* And set the start address. */
thread_table[0].data.registers.integer_registers.rip =
prepare_process_ret_val.first_instruction_address;
/* Finally we set the current thread. */
cpu_private_data.thread_index = 0;
}
/* Set up the timer hardware to generate interrupts 200 times a second. */
outb(0x43, 0x36);
outb(0x40, 78);
outb(0x40, 23);
/* Set up the keyboard controller. */
/* Empty the keyboard buffer. */
{
register unsigned char status_byte;
do
{
status_byte=inb(0x64);
if ((status_byte&3)==1)
{
inb(0x60);
}
} while((status_byte&0x3)!=0x0);
}
/* Change the command byte to enable interrupts. */
outb(0x64, 0x20);
{
register unsigned char keyboard_controller_command_byte;
{
register unsigned char status_byte;
do
{
status_byte=inb(0x64);
} while((status_byte&3)!=1);
}
keyboard_controller_command_byte=inb(0x60);
/* Enable keyboard interrupts. */
keyboard_controller_command_byte|=1;
kprints("Keyboard controller command byte:");
kprinthex(keyboard_controller_command_byte);
kprints("\n");
outb(0x64, 0x60);
outb(0x60, keyboard_controller_command_byte);
/* Wait until command is done. */
{
register unsigned char status_byte;
do
{
status_byte=inb(0x64);
} while((status_byte&0x2)!=0x0);
}
}
/* Now we set up the interrupt controller to allow timer and keyboard
interrupts. */
outb(0x20, 0x11);
outb(0xA0, 0x11);
outb(0x21, 0x20);
outb(0xA1, 0x28);
outb(0x21, 1<<2);
outb(0xA1, 2);
outb(0x21, 1);
outb(0xA1, 1);
outb(0x21, 0xfc);
outb(0xA1, 0xff);
clear_screen();
kprints("\n\n\nThe kernel has booted!\n\n\n");
/* Now go back to the assembly language code and let the process run. */
}
