1 MINIX3 SYSTEM TASK MINIX3 SYSTEM TASK Cátedra: Diseño e Implementación de Sistemas Operativos UTN-FRSF Tomado de: Operating Systems Design and Implementation, Third Edition

2 Generalidades Como consecuencias de que la mayor cantidad de componentes están fuera del kernel como procesos independientes, ellos están inhibidos de realizar ciertas operaciones privilegidadas.  Manipulación de Tablas del Kernel  Operaciones de Entrada/Salida  Copias de areas de memoria entre procesos La solución es que el kernel ofrezca esos servicios a Drivers y Servers (no a procesos de usuario ordinarios). Estos servicios se brindan a traves de un proceso que se llama SYSTEM TASK (SYSTASK) que es un representante del kernel. SYSTASK se compila en el mismo binario que el kernel y accede a todas sus variables y funciones.

3 Generalidades Un ejemplo de uso de la SYSTASK por parte de un Device Driver en modo usuario es el uso de la Kernel call sys_irqctl para instalar un handler de interrupción. En realidad lo que hace la SYSTASK es instalar un generic_handler que se encarga de notificar a la tarea correspondiente del Device Driver. El programa main() de la SYSTASK funciona en forma similar al resto de las tareas. Es un loop que espera una petición a través de un mensaje, despacha el procedimiento correspondiente y retorna una respuesta. El programa main() se encuentra en system.c del directorio del kernel. El resto de los procedimientos en el directorio system.

4 Mensajes a SYSTASK Message type FromMeaning sys_forkPMA process has forked sys_execPMSet stack pointer after EXEC call sys_exitPMA process has exited sys_nicePMSet scheduling priority sys_privctlRSSet or change privileges sys_tracePM Carry out an operation of the PTRACE call sys_killPM, FS, TTYSend signal to a process after KILL call sys_getksigPMPM is checking for pending signals sys_endksigPMPM has finished processing signal

5 Mensajes a SYSTASK Message typeFromMeaning sys_sigsendPMSend a signal to a process sys_sigreturnPMCleanup after completion of a signal sys_irqctlDriversEnable, disable, or configure interrupt sys_devioDriversRead from or write to an I/O port sys_sdevioDriversRead or write string from/to I/O port sys_vdevioDriversCarry out a vector of I/O requests sys_int86DriversDo a real-mode BIOS call sys_newmapPMSet up a process memory map sys_segctlDriversAdd segment and get selector (far data access)

6 Mensajes a SYSTASK Message typeFromMeaning sys_memsetPMWrite char to memory area sys_umapDriversConvert virtual address to physical address sys_vircopyFS, DriversCopy using pure virtual addressing sys_physcopyDriversCopy using physical addressing sys_virvcopyAnyVector of VCOPY requests sys_physvcopyAnyVector of PHYSCOPY requests sys_timesPMGet uptime and process times sys_setalarmPM, FS, DriversSchedule a synchronous alarm sys_abortPM, TTYPanic: MINIX is unable to continue sys_getinfoAnyRequest system information

7 Mensajes a SYSTASK Los Kernel Calls sys_irqctl, sys_devio, sys_sdevio, y sys_vdevio son provistos por MINIX3 para soportar los Device Drivers en espacio de usuario. sys_newmap es invocado por el PM cuando ha cambiado la memoria asignada al proceso, por lo tanto el Kernel debe actualizar sus tablas. Sys_segctl y sys_memset brindan una forma segura de que el proceso pueda acceder a areas de memoria fuera de su espacio de usuario. El area de memoria desde 0xa0000 to 0xfffff se reserva para dispositivos de E/S. Sys_segctl es usaso por los Device Drivers para obtener el Selector del Segmento que le permitirá acceder a ese rango de memoria.

8 Mensajes a SYSTASK sys_memset se usa para escribir datos en un area de memoria que no le pertenece al proceso, como por ejemplo para poner Ceros en ella. Sys_umap convierte direcciones virtuales en direcciones físicas. Sys_vircopy y sys_physcopy copian regiones de memoria utilizando direccionamiento virtual o físico. sys_virvcopy y sys_physvcopy son las versiones vectorizadas de las previsas. sys_getinfo maneja un amplio rango de pedidos de información al Kernel.

9 Implementacion de SYSTASK 09753 /*===========================================================================* 09754 * sys_task * 09755 *===========================================================================*/ 09756 PUBLIC void sys_task() 09757 { 09758 /* Main entry point of sys_task. Get the message and dispatch on type. */ 09759 static message m; 09760 register int result; 09761 register struct proc *caller_ptr; 09762 unsigned int call_nr; 09763 int s; 09765 /* Initialize the system task. */ 09766 initialize(); 09768 while (TRUE) { 09769 /* Get work. Block and wait until a request message arrives. */ 09770 receive(ANY, &m); 09771 call_nr = (unsigned) m.m_type - KERNEL_CALL; 09772 caller_ptr = proc_addr(m.m_source); 09773 09774 /* See if the caller made a valid request and try to handle it. */ 09775 if (! (priv(caller_ptr)->s_call_mask & (1

10 Implementacion de SYSTASK 09786 /* Send a reply, unless inhibited by a handler function. Use the kernel 09787 * function lock_send() to prevent a system call trap. The destination 09788 * is known to be blocked waiting for a message. 09789 */ 09790 if (result != EDONTREPLY) { 09791 m.m_type = result; /* report status of call */ 09792 if (OK != (s=lock_send(m.m_source, &m))) { 09793 kprintf("SYSTEM, reply to %d failed: %d\n", m.m_source, s); 09794 } 09795 } 09796 } 09797 }

11 Implementacion de SYSTASK 09747 #define map(call_nr, handler) \ 09748 {extern int dummy[NR_SYS_CALLS>(unsigned)(call_nr-KERNEL_CALL) ? 1:-1];} \ 09749 call_vec[(call_nr-KERNEL_CALL)] = (handler) 09799 /*===========================================================================* 09800 * initialize * 09801 *===========================================================================*/ 09802 PRIVATE void initialize(void) 09803 { 09804 register struct priv *sp; 09805 int i; 09806 09807 /* Initialize IRQ handler hooks. Mark all hooks available. */ 09808 for (i=0; is_alarm_timer)); 09815 } 09816 09817 /* Initialize the call vector to a safe default handler. Some kernel calls 09818 * may be disabled or nonexistant. Then explicitly map known calls to their 09819 * handler functions. This is done with a macro that gives a compile error 09820 * if an illegal call number is used. The ordering is not important here. 09821 */ 09822 for (i=0; i(unsigned)(call_nr-KERNEL_CALL) .", "description": "1:-1];} \ 09749 call_vec[(call_nr-KERNEL_CALL)] = (handler) 09799 /*===========================================================================* 09800 * initialize * 09801 *===========================================================================*/ 09802 PRIVATE void initialize(void) 09803 { 09804 register struct priv *sp; 09805 int i; 09806 09807 /* Initialize IRQ handler hooks. Mark all hooks available. */ 09808 for (i=0; is_alarm_timer)); 09815 } 09816 09817 /* Initialize the call vector to a safe default handler. Some kernel calls 09818 * may be disabled or nonexistant. Then explicitly map known calls to their 09819 * handler functions. This is done with a macro that gives a compile error 09820 * if an illegal call number is used. The ordering is not important here. 09821 */ 09822 for (i=0; i(unsigned)(call_nr-KERNEL_CALL) ." title="1:-1];} \ 09749 call_vec[(call_nr-KERNEL_CALL)] = (handler) 09799 /*===========================================================================* 09800 * initialize * 09801 *===========================================================================*/ 09802 PRIVATE void initialize(void) 09803 { 09804 register struct priv *sp; 09805 int i; 09806 09807 /* Initialize IRQ handler hooks. Mark all hooks available. */ 09808 for (i=0; is_alarm_timer)); 09815 } 09816 09817 /* Initialize the call vector to a safe default handler. Some kernel calls 09818 * may be disabled or nonexistant. Then explicitly map known calls to their 09819 * handler functions. This is done with a macro that gives a compile error 09820 * if an illegal call number is used. The ordering is not important here. 09821 */ 09822 for (i=0; iFS read 2.FS->DD Si no esta en cache rw_block() 3.DD->SYSTASK sys_devio 4.SYSTASK->DD ACK 5.SYSTASK->DD INTERRUPT 6.DD->SYSTASK sys_copy from kernel to FS cache 7.SYSTASK->DD ACK 8.DD->FS ACK 9.FS->SYSTASK sys_copy from FS cache to user buffer 10.SYSTASK->FS ACK 11.FS->USER ACK" class="image_link uk-text-large uk-margin-small-left uk-margin-small-right"> 14 Impacto en Performance 1.USER->FS read 2.FS->DD Si no esta en cache rw_block() 3.DD->SYSTASK sys_devio 4.SYSTASK->DD ACK 5.SYSTASK->DD INTERRUPT 6.DD->SYSTASK sys_copy from kernel to FS cache 7.SYSTASK->DD ACK 8.DD->FS ACK 9.FS->SYSTASK sys_copy from FS cache to user buffer 10.SYSTASK->FS ACK 11.FS->USER ACK { "@context": "http://schema.org", "@type": "ImageObject", "contentUrl": "http://images.slideplayer.es/14/4325006/slides/slide_14.jpg", "name": "Impacto en Performance 1.USER->FS read 2.FS->DD Si no esta en cache rw_block() 3.DD->SYSTASK sys_devio 4.SYSTASK->DD ACK 5.SYSTASK->DD INTERRUPT 6.DD->SYSTASK sys_copy from kernel to FS cache 7.SYSTASK->DD ACK 8.DD->FS ACK 9.FS->SYSTASK sys_copy from FS cache to user buffer 10.SYSTASK->FS ACK 11.FS->USER ACK", "description": "Impacto en Performance 1.USER->FS read 2.FS->DD Si no esta en cache rw_block() 3.DD->SYSTASK sys_devio 4.SYSTASK->DD ACK 5.SYSTASK->DD INTERRUPT 6.DD->SYSTASK sys_copy from kernel to FS cache 7.SYSTASK->DD ACK 8.DD->FS ACK 9.FS->SYSTASK sys_copy from FS cache to user buffer 10.SYSTASK->FS ACK 11.FS->USER ACK", "width": "800" } 15 Impacto en Performance 1.USER->FS read 2.FS->SYSTASK sys_copy from FS cache to user buffer 3.SYSTASK->FS ACK 4.FS->USER ACK

12 Implementacion de SYSTASK 09826 /* Process management. */ 09827 map(SYS_FORK, do_fork); /* a process forked a new process */ 09828 map(SYS_EXEC, do_exec); /* update process after execute */ 09829 map(SYS_EXIT, do_exit); /* clean up after process exit */ 09830 map(SYS_NICE, do_nice); /* set scheduling priority */ 09831 map(SYS_PRIVCTL, do_privctl); /* system privileges control */ 09832 map(SYS_TRACE, do_trace); /* request a trace operation */ 09833 09834 /* Signal handling. */ 09835 map(SYS_KILL, do_kill); /* cause a process to be signaled */ 09836 map(SYS_GETKSIG, do_getksig); /* PM checks for pending signals */ 09837 map(SYS_ENDKSIG, do_endksig); /* PM finished processing signal */ 09838 map(SYS_SIGSEND, do_sigsend); /* start POSIX-style signal */ 09839 map(SYS_SIGRETURN, do_sigreturn); /* return from POSIX-style signal */ 09840 09841 /* Device I/O. */ 09842 map(SYS_IRQCTL, do_irqctl); /* interrupt control operations */ 09843 map(SYS_DEVIO, do_devio); /* inb, inw, inl, outb, outw, outl */ 09844 map(SYS_SDEVIO, do_sdevio); /* phys_insb, _insw, _outsb, _outsw */ 09845 map(SYS_VDEVIO, do_vdevio); /* vector with devio requests */ 09846 map(SYS_INT86, do_int86); /* real-mode BIOS calls */ 09847 09848 /* Memory management. */ 09849 map(SYS_NEWMAP, do_newmap); /* set up a process memory map */ 09850 map(SYS_SEGCTL, do_segctl); /* add segment and get selector */ 09851 map(SYS_MEMSET, do_memset); /* write char to memory area */

13 Implementacion de SYSTASK 09853 /* Copying. */ 09854 map(SYS_UMAP, do_umap); /* map virtual to physical address */ 09855 map(SYS_VIRCOPY, do_vircopy); /* use pure virtual addressing */ 09856 map(SYS_PHYSCOPY, do_physcopy); /* use physical addressing */ 09857 map(SYS_VIRVCOPY, do_virvcopy); /* vector with copy requests */ 09858 map(SYS_PHYSVCOPY, do_physvcopy); /* vector with copy requests */ 09859 09860 /* Clock functionality. */ 09861 map(SYS_TIMES, do_times); /* get uptime and process times */ 09862 map(SYS_SETALARM, do_setalarm); /* schedule a synchronous alarm */ 09863 09864 /* System control. */ 09865 map(SYS_ABORT, do_abort); /* abort MINIX */ 09866 map(SYS_GETINFO, do_getinfo); /* request system information */ 09867 }

14 Impacto en Performance 1.USER->FS read 2.FS->DD Si no esta en cache rw_block() 3.DD->SYSTASK sys_devio 4.SYSTASK->DD ACK 5.SYSTASK->DD INTERRUPT 6.DD->SYSTASK sys_copy from kernel to FS cache 7.SYSTASK->DD ACK 8.DD->FS ACK 9.FS->SYSTASK sys_copy from FS cache to user buffer 10.SYSTASK->FS ACK 11.FS->USER ACK

15 Impacto en Performance 1.USER->FS read 2.FS->SYSTASK sys_copy from FS cache to user buffer 3.SYSTASK->FS ACK 4.FS->USER ACK