diff options
Diffstat (limited to 'sys/kern')
| -rw-r--r-- | sys/kern/kern_filedesc.c | 123 | ||||
| -rw-r--r-- | sys/kern/kern_loader.c | 33 | ||||
| -rw-r--r-- | sys/kern/kern_sched.c | 275 | ||||
| -rw-r--r-- | sys/kern/kern_syscall.c | 44 | ||||
| -rw-r--r-- | sys/kern/vfs_init.c | 5 | ||||
| -rw-r--r-- | sys/kern/vfs_lookup.c | 61 |
6 files changed, 474 insertions, 67 deletions
diff --git a/sys/kern/kern_filedesc.c b/sys/kern/kern_filedesc.c new file mode 100644 index 0000000..a943714 --- /dev/null +++ b/sys/kern/kern_filedesc.c @@ -0,0 +1,123 @@ +/* + * Copyright (c) 2023-2024 Ian Marco Moffett and the Osmora Team. + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * 1. Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * 3. Neither the name of Hyra nor the names of its + * contributors may be used to endorse or promote products derived from + * this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" + * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE + * POSSIBILITY OF SUCH DAMAGE. + */ + +#include <sys/filedesc.h> +#include <sys/proc.h> +#include <sys/sched.h> +#include <vm/dynalloc.h> +#include <assert.h> +#include <string.h> + +/* + * Allocate a file descriptor. + * + * @td: Thread to allocate from, NULL for current thread. + */ +struct filedesc * +fd_alloc(struct proc *td) +{ + struct filedesc *fd; + + if (td == NULL) { + td = this_td(); + __assert(td != NULL); + } + + for (size_t i = 0; i < PROC_MAX_FDS; ++i) { + if (td->fds[i] != NULL) { + continue; + } + + fd = dynalloc(sizeof(struct filedesc)); + memset(fd, 0, sizeof(struct filedesc)); + + if (fd == NULL) { + return NULL; + } + + fd->fdno = i; + td->fds[i] = fd; + return fd; + } + + return NULL; +} + +/* + * Fetch a file descriptor from a file descriptor + * number. + * + * @td: Thread to fetch from, NULL for current thread. + * @fdno: File descriptor to fetch + */ +struct filedesc * +fd_from_fdnum(const struct proc *td, int fdno) +{ + if (td == NULL) { + td = this_td(); + __assert(td != NULL); + } + + if (fdno < 0 || fdno > PROC_MAX_FDS) { + return NULL; + } + + for (size_t i = 0; i < PROC_MAX_FDS; ++i) { + if (i == fdno && td->fds[i] != NULL) { + return td->fds[i]; + } + } + + return NULL; +} + +/* + * Close a file descriptor from its fd number. + * + * @td: Thread to fetch from, NULL for current thread. + * @fdno: File descriptor number to close. + */ +void +fd_close_fdnum(struct proc *td, int fdno) +{ + struct filedesc *fd; + + if (td == NULL) { + td = this_td(); + __assert(td != NULL); + } + + fd = fd_from_fdnum(td, fdno); + if (fd == NULL) { + return; + } + + dynfree(fd); + td->fds[fdno] = NULL; +} diff --git a/sys/kern/kern_loader.c b/sys/kern/kern_loader.c index 53ba8c2..23249ea 100644 --- a/sys/kern/kern_loader.c +++ b/sys/kern/kern_loader.c @@ -61,12 +61,10 @@ int loader_load(struct vas vas, const void *dataptr, struct auxval *auxv, vm_prot_t prot = PROT_USER; uintptr_t physmem; - uintptr_t max_addr, map_addr; - size_t misalign, page_count; - + size_t misalign, page_count, map_len; int status; - const size_t GRANULE = vm_get_page_size(); + const size_t GRANULE = vm_get_page_size(); void *tmp_ptr; if (auxv == NULL) { @@ -93,26 +91,8 @@ int loader_load(struct vas vas, const void *dataptr, struct auxval *auxv, misalign = phdr->p_vaddr & (GRANULE - 1); page_count = __DIV_ROUNDUP(phdr->p_memsz + misalign, GRANULE); - max_addr = phdr->p_vaddr + (GRANULE * page_count); - - /* - * We are assuming this is a user program that we are loading. - * All user programs should be on the lower half of the address - * space. We will check that before we begin doing anything here. - * - * We are also going to check if the virtual address the program - * header refers to overflows into the higher half. If anything - * goes into the higher half, we won't simply drop the phdr, - * we'll instead assume caller error and return -EINVAL. - */ - if (phdr->p_vaddr >= VM_HIGHER_HALF) { - return -EINVAL; - } else if (max_addr >= VM_HIGHER_HALF) { - /* Overflows into higher half */ - return -EINVAL; - } - physmem = vm_alloc_pageframe(page_count); + map_len = page_count * GRANULE; /* Do we not have enough page frames? */ if (physmem == 0) { @@ -121,14 +101,9 @@ int loader_load(struct vas vas, const void *dataptr, struct auxval *auxv, return -ENOMEM; } - map_addr = phdr->p_vaddr + load_base; - status = vm_map_create(vas, map_addr, physmem, prot, page_count*GRANULE); + status = vm_map_create(vas, phdr->p_vaddr + load_base, physmem, prot, map_len); if (status != 0) { - DBG("Failed to map 0x%p - 0x%p\n", - phdr->p_vaddr + load_base, - (phdr->p_vaddr + load_base) + (page_count * GRANULE)); - return status; } diff --git a/sys/kern/kern_sched.c b/sys/kern/kern_sched.c index f61c592..7b3776d 100644 --- a/sys/kern/kern_sched.c +++ b/sys/kern/kern_sched.c @@ -34,10 +34,36 @@ #include <sys/timer.h> #include <sys/cdefs.h> #include <sys/spinlock.h> +#include <sys/loader.h> +#include <sys/panic.h> +#include <sys/machdep.h> +#include <sys/filedesc.h> +#include <fs/initramfs.h> #include <vm/dynalloc.h> +#include <vm/physseg.h> +#include <vm/pmap.h> +#include <vm/map.h> +#include <vm/vm.h> #include <assert.h> #include <string.h> +#define STACK_PAGES 8 +#define STACK_SIZE (STACK_PAGES*vm_get_page_size()) + +/* + * The PHYS_TO_VIRT/VIRT_TO_PHYS macros convert + * addresses to lower and higher half addresses. + * Userspace addresses are on the lower half, + * therefore, we can just wrap over these to + * keep things simple. + * + * XXX: TODO: This won't work when not identity mapping + * lowerhalf addresses. Once that is updated, + * get rid of this. + */ +#define USER_TO_KERN(user) PHYS_TO_VIRT(user) +#define KERN_TO_USER(kern) VIRT_TO_PHYS(kern) + /* * Thread ready queue - all threads ready to be * scheduled should be added to this queue. @@ -51,9 +77,6 @@ static size_t nthread = 0; */ static struct spinlock tdq_lock = {0}; -/* In sys/<machine>/<machine>/switch.S */ -void __sched_switch_to(struct trapframe *tf); - static inline void sched_oneshot(void) { @@ -97,13 +120,11 @@ sched_dequeue_td(void) spinlock_acquire(&tdq_lock); - if (TAILQ_EMPTY(&td_queue)) { - goto done; + if (!TAILQ_EMPTY(&td_queue)) { + td = TAILQ_FIRST(&td_queue); + TAILQ_REMOVE(&td_queue, td, link); } - td = TAILQ_FIRST(&td_queue); - TAILQ_REMOVE(&td_queue, td, link); -done: spinlock_release(&tdq_lock); return td; } @@ -121,12 +142,101 @@ sched_enter(void) } } +static uintptr_t +sched_init_stack(void *stack_top, char *argvp[], char *envp[], struct auxval auxv) +{ + uintptr_t *sp = stack_top; + void *env_ptr = NULL, *argv_ptr = NULL; + size_t argc, envc, len; + + /* Copy argument and environment strings */ + for (envc = 0; envp[envc] != NULL; ++envc) { + len = strlen(envp[envc]); + sp -= len - 1; + memcpy(sp, envp[envc], len); + } + + __assert(envc >= 1); + env_ptr = sp; + + for (argc = 0; argvp[argc] != NULL; ++argc) { + len = strlen(argvp[argc]); + sp -= len - 1; + memcpy(sp, argvp[argc], len); + } + + __assert(argc >= 1); + argv_ptr = sp; + + /* Ensure the stack is aligned */ + sp = (void *)__ALIGN_DOWN((uintptr_t)sp, 16); + if (((argc + envc + 1) & 1) != 0) + --sp; + + AUXVAL(sp, AT_NULL, 0x0); + AUXVAL(sp, AT_SECURE, 0x0); + AUXVAL(sp, AT_ENTRY, auxv.at_entry); + AUXVAL(sp, AT_PHDR, auxv.at_phdr); + AUXVAL(sp, AT_PHNUM, auxv.at_phnum); + STACK_PUSH(sp, 0); + + /* Push environment string pointers */ + for (int i = 0; i < envc; ++i) { + len = strlen(env_ptr); + sp -= len; + + *sp = (uintptr_t)KERN_TO_USER((uintptr_t)env_ptr); + env_ptr = (char *)env_ptr + len; + } + + /* Push argument string pointers */ + STACK_PUSH(sp, 0); + for (int i = 0; i < argc; ++i) { + len = strlen(argv_ptr); + sp -= len; + + *sp = (uintptr_t)KERN_TO_USER((uintptr_t)argv_ptr); + argv_ptr = (char *)argv_ptr + len; + } + + STACK_PUSH(sp, argc); + + return (uintptr_t)sp; +} + +static uintptr_t +sched_create_stack(struct vas vas, bool user, char *argvp[], + char *envp[], struct auxval auxv, struct proc *td) +{ + int status; + uintptr_t stack; + const vm_prot_t USER_STACK_PROT = PROT_WRITE | PROT_USER; + + if (!user) { + stack = (uintptr_t)dynalloc(STACK_SIZE); + td->stack_base = (uintptr_t)stack; + return sched_init_stack((void *)(stack + STACK_SIZE), argvp, envp, auxv); + } + + stack = vm_alloc_pageframe(STACK_PAGES); + td->stack_base = stack; + status = vm_map_create(vas, stack, stack, USER_STACK_PROT, STACK_SIZE); + + if (status != 0) { + return 0; + } + + memset(USER_TO_KERN(stack), 0, STACK_SIZE); + stack = sched_init_stack((void *)USER_TO_KERN(stack + STACK_SIZE), argvp, envp, auxv); + return stack; +} + static struct proc * -sched_create_td(uintptr_t rip) +sched_create_td(uintptr_t rip, char *argvp[], char *envp[], struct auxval auxv, + struct vas vas, bool is_user) { - const size_t STACK_SIZE = 0x100000; /* 1 MiB */ struct proc *td; - void *stack; + uintptr_t stack; struct trapframe *tf; tf = dynalloc(sizeof(struct trapframe)); @@ -134,32 +244,103 @@ sched_create_td(uintptr_t rip) return NULL; } - stack = dynalloc(STACK_SIZE); - if (stack == NULL) { + td = dynalloc(sizeof(struct proc)); + if (td == NULL) { + /* TODO: Free stack */ dynfree(tf); return NULL; } - td = dynalloc(sizeof(struct proc)); - if (td == NULL) { + stack = sched_create_stack(vas, is_user, argvp, envp, auxv, td); + if (stack == 0) { dynfree(tf); - dynfree(stack); + dynfree(td); return NULL; } memset(tf, 0, sizeof(struct trapframe)); - memset(stack, 0, STACK_SIZE); + memset(td, 0, sizeof(struct proc)); /* Setup process itself */ td->pid = 0; /* Don't assign PID until enqueued */ td->cpu = NULL; /* Not yet assigned a core */ td->tf = tf; + td->addrsp = vas; + td->is_user = is_user; + processor_init_pcb(td); + + /* Allocate standard file descriptors */ + __assert(fd_alloc(td) != NULL); /* STDIN */ + __assert(fd_alloc(td) != NULL); /* STDOUT */ + __assert(fd_alloc(td) != NULL); /* STDERR */ /* Setup trapframe */ - init_frame(tf, rip, (uintptr_t)stack + STACK_SIZE - 1); + if (!is_user) { + init_frame(tf, rip, (uintptr_t)stack); + } else { + init_frame_user(tf, rip, KERN_TO_USER(stack)); + } return td; } +static void +sched_destroy_td(struct proc *td) +{ + processor_free_pcb(td); + + /* + * User stacks are allocated with vm_alloc_pageframe(), + * while kernel stacks are allocated with dynalloc(). + * We want to check if we are a user program or kernel + * program to perform the proper deallocation method. + */ + if (td->is_user) { + vm_free_pageframe(td->stack_base, STACK_PAGES); + } else { + dynfree((void *)td->stack_base); + } + + /* Close all of the file descriptors */ + for (size_t i = 0; i < PROC_MAX_FDS; ++i) { + fd_close_fdnum(td, i); + } + + pmap_free_vas(vm_get_ctx(), td->addrsp); + dynfree(td); +} + +void +sched_exit(void) +{ + struct proc *td; + struct vas kvas = vm_get_kvas(); + + intr_mask(); + + td = this_td(); + + /* Switch back to the kernel address space and destroy ourself */ + pmap_switch_vas(vm_get_ctx(), kvas); + sched_destroy_td(td); + + intr_unmask(); + sched_enter(); +} + +/* + * Get the current running thread. + */ +struct proc * +this_td(void) +{ + struct sched_state *state; + struct cpu_info *ci; + + ci = this_cpu(); + state = &ci->sched_state; + return state->td; +} + /* * Thread context switch routine */ @@ -174,43 +355,73 @@ sched_context_switch(struct trapframe *tf) * If we have no threads, we should not * preempt at all. */ - if (nthread == 0) { - goto done; - } else if ((next_td = sched_dequeue_td()) == NULL) { - /* Empty */ - goto done; + if (nthread == 0 || (next_td = sched_dequeue_td()) == NULL) { + sched_oneshot(); + return; } + /* + * If we have a thread currently running and we are switching + * to another, we shall save our current register state + * by copying the trapframe. + */ if (state->td != NULL) { - /* Save our trapframe */ td = state->td; memcpy(td->tf, tf, sizeof(struct trapframe)); } - /* Copy to stack */ + /* Copy over the next thread's register state to us */ memcpy(tf, next_td->tf, sizeof(struct trapframe)); td = state->td; state->td = next_td; + /* Re-enqueue the previous thread if it exists */ if (td != NULL) { sched_enqueue_td(td); } -done: + + /* Do architecture specific context switch logic */ + processor_switch_to(td, next_td); + + /* Done, switch out our vas and oneshot */ + pmap_switch_vas(vm_get_ctx(), next_td->addrsp); sched_oneshot(); } void sched_init(void) { + struct proc *init; + struct auxval auxv = {0}, ld_auxv = {0}; + struct vas vas = pmap_create_vas(vm_get_ctx()); + const char *init_bin, *ld_bin; + + char *ld_path; + char *argv[] = {"/usr/sbin/init", NULL}; + char *envp[] = {"", NULL}; + TAILQ_INIT(&td_queue); - /* - * TODO: Create init with sched_create_td() - * and enqueue with sched_enqueue_td() - */ - (void)sched_create_td; - (void)sched_enqueue_td; + if ((init_bin = initramfs_open("/usr/sbin/init")) == NULL) { + panic("Could not open /usr/boot/init\n"); + } + if (loader_load(vas, init_bin, &auxv, 0, &ld_path) != 0) { + panic("Could not load init\n"); + } + if ((ld_bin = initramfs_open(ld_path)) == NULL) { + panic("Could not open %s\n", ld_path); + } + if (loader_load(vas, ld_bin, &ld_auxv, 0x00, NULL) != 0) { + panic("Could not load %s\n", ld_path); + } + + init = sched_create_td((uintptr_t)ld_auxv.at_entry, argv, envp, ld_auxv, vas, true); + if (init == NULL) { + panic("Failed to create thread for init\n"); + } + + sched_enqueue_td(init); } /* diff --git a/sys/kern/kern_syscall.c b/sys/kern/kern_syscall.c new file mode 100644 index 0000000..b6e31d1 --- /dev/null +++ b/sys/kern/kern_syscall.c @@ -0,0 +1,44 @@ +/* + * Copyright (c) 2023-2024 Ian Marco Moffett and the Osmora Team. + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * 1. Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * 3. Neither the name of Hyra nor the names of its + * contributors may be used to endorse or promote products derived from + * this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" + * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE + * POSSIBILITY OF SUCH DAMAGE. + */ + +#include <sys/syscall.h> +#include <sys/sched.h> +#include <sys/cdefs.h> +#include <sys/types.h> + +__noreturn static uint64_t +sys_exit(struct syscall_args *args) +{ + sched_exit(); + __builtin_unreachable(); +} + +uint64_t(*g_syscall_table[__MAX_SYSCALLS])(struct syscall_args *args) = { + sys_exit, +}; diff --git a/sys/kern/vfs_init.c b/sys/kern/vfs_init.c index 18ab3ac..0d8d194 100644 --- a/sys/kern/vfs_init.c +++ b/sys/kern/vfs_init.c @@ -40,8 +40,10 @@ __MODULE_NAME("vfs"); __KERNEL_META("$Hyra$: vfs.c, Ian Marco Moffett, " "Hyra Virtual File System"); +#define INITRAMFS_ID 0 + static struct fs_info filesystems[] = { - { "initramfs", &g_initramfs_ops } + [INITRAMFS_ID] = { "initramfs", &g_initramfs_ops } }; struct vnode *g_root_vnode = NULL; @@ -76,4 +78,5 @@ vfs_init(void) } g_root_vnode->vops = &g_initramfs_vops; + g_root_vnode->fs = &filesystems[INITRAMFS_ID]; } diff --git a/sys/kern/vfs_lookup.c b/sys/kern/vfs_lookup.c index 748d5e2..1398964 100644 --- a/sys/kern/vfs_lookup.c +++ b/sys/kern/vfs_lookup.c @@ -29,11 +29,13 @@ #include <sys/types.h> #include <sys/vfs.h> +#include <sys/mount.h> +#include <sys/errno.h> #include <vm/dynalloc.h> #include <string.h> /* - * Fetches the filename within a past at + * Fetches the filename within a path at * the nth index denoted by `idx' * * Returns memory allocated by dynalloc() @@ -115,9 +117,58 @@ vfs_get_fname_at(const char *path, size_t idx) return ret; } -struct vnode * -vfs_path_to_node(const char *path) +/* + * Fetches a vnode from a path. + * + * @path: Path to fetch vnode from. + * @vp: Output var for fetched vnode. + * + * Returns 0 on success. + */ +int +vfs_path_to_node(const char *path, struct vnode **vp) { - /* TODO */ - return NULL; + struct vnode *vnode = g_root_vnode; + struct fs_info *fs; + char *name; + int s = 0, fs_caps = 0; + + if (strcmp(path, "/") == 0 || !vfs_is_valid_path(path)) { + return -1; + } else if (*path != '/') { + return -1; + } + + /* Fetch filesystem capabilities if we can */ + if (vnode->fs != NULL) { + fs = vnode->fs; + fs_caps = fs->caps; + } + + /* + * If the filesystem requires full-path lookups, we can try + * throwing the full path at the filesystem to see if + * it'll give us a vnode. + */ + if (__TEST(fs_caps, FSCAP_FULLPATH)) { + s = vfs_vget(g_root_vnode, path, &vnode); + goto done; + } + + for (size_t i = 0;; ++i) { + name = vfs_get_fname_at(path, i); + if (name == NULL) break; + + s = vfs_vget(vnode, name, &vnode); + dynfree(name); + + if (s != 0) break; + } + +done: + if (vp != NULL && s == 0) { + *vp = vnode; + } + + return s; } |