summaryrefslogtreecommitdiff
path: root/sys
diff options
context:
space:
mode:
Diffstat (limited to 'sys')
-rw-r--r--sys/arch/amd64/amd64/machdep.c82
-rw-r--r--sys/arch/amd64/amd64/pmap.c13
-rw-r--r--sys/arch/amd64/amd64/spectre.c7
-rw-r--r--sys/arch/amd64/amd64/syscall.S43
-rw-r--r--sys/arch/amd64/amd64/syscall.c49
-rw-r--r--sys/arch/amd64/amd64/trap.S10
-rw-r--r--sys/arch/amd64/amd64/trap.c12
-rw-r--r--sys/fs/initramfs.c9
-rw-r--r--sys/include/arch/amd64/cpu.h40
-rw-r--r--sys/include/arch/amd64/frame.h7
-rw-r--r--sys/include/arch/amd64/pcb.h39
-rw-r--r--sys/include/arch/amd64/trap.h2
-rw-r--r--sys/include/sys/filedesc.h53
-rw-r--r--sys/include/sys/loader.h27
-rw-r--r--sys/include/sys/machdep.h4
-rw-r--r--sys/include/sys/mount.h7
-rw-r--r--sys/include/sys/proc.h10
-rw-r--r--sys/include/sys/sched.h2
-rw-r--r--sys/include/sys/syscall.h56
-rw-r--r--sys/include/sys/vfs.h2
-rw-r--r--sys/include/sys/vnode.h1
-rw-r--r--sys/include/vm/pmap.h6
-rw-r--r--sys/include/vm/vm.h1
-rw-r--r--sys/kern/kern_filedesc.c123
-rw-r--r--sys/kern/kern_loader.c33
-rw-r--r--sys/kern/kern_sched.c275
-rw-r--r--sys/kern/kern_syscall.c44
-rw-r--r--sys/kern/vfs_init.c5
-rw-r--r--sys/kern/vfs_lookup.c61
-rw-r--r--sys/vm/vm_init.c9
-rw-r--r--sys/vm/vm_map.c6
31 files changed, 948 insertions, 90 deletions
diff --git a/sys/arch/amd64/amd64/machdep.c b/sys/arch/amd64/amd64/machdep.c
index 6342aab..47d6dd0 100644
--- a/sys/arch/amd64/amd64/machdep.c
+++ b/sys/arch/amd64/amd64/machdep.c
@@ -40,6 +40,7 @@
#include <machine/spectre.h>
#include <machine/cpu.h>
#include <machine/uart.h>
+#include <machine/cpuid.h>
#include <vm/vm.h>
#include <vm/dynalloc.h>
#include <vm/physseg.h>
@@ -56,6 +57,8 @@ __KERNEL_META("$Hyra$: machdep.c, Ian Marco Moffett, "
#define INIT_FLAG_IOAPIC 0x00000001U
#define INIT_FLAG_ACPI 0x00000002U
+void syscall_isr(void);
+
static inline void
init_tss(struct cpu_info *cur_cpu)
{
@@ -78,18 +81,28 @@ interrupts_init(void)
idt_set_desc(0x8, IDT_TRAP_GATE_FLAGS, ISR(double_fault), 0);
idt_set_desc(0xA, IDT_TRAP_GATE_FLAGS, ISR(invl_tss), 0);
idt_set_desc(0xB, IDT_TRAP_GATE_FLAGS, ISR(segnp), 0);
+ idt_set_desc(0xC, IDT_TRAP_GATE_FLAGS, ISR(ss_fault), 0);
idt_set_desc(0xD, IDT_TRAP_GATE_FLAGS, ISR(general_prot), 0);
idt_set_desc(0xE, IDT_TRAP_GATE_FLAGS, ISR(page_fault), 0);
+ idt_set_desc(0x80, IDT_INT_GATE_USER, ISR(syscall_isr), 0);
idt_load();
}
+static bool
+is_sse_supported(void)
+{
+ uint32_t edx, unused;
+
+ __CPUID(0x00000001, unused, unused, unused, edx);
+ return __TEST(edx, __BIT(25)) && __TEST(edx, __BIT(26));
+}
+
void
processor_halt(void)
{
__ASMV("cli; hlt");
}
-
/*
* Send char to serial for debugging purposes.
*/
@@ -144,11 +157,63 @@ intr_unmask(void)
__ASMV("sti");
}
+int
+processor_init_pcb(struct proc *proc)
+{
+ struct pcb *pcb = &proc->pcb;
+ const uint16_t FPU_FCW = 0x33F;
+ const uint32_t SSE_MXCSR = 0x1F80;
+
+ /* Allocate FPU save area, aligned on a 16 byte boundary */
+ pcb->fpu_state = PHYS_TO_VIRT(vm_alloc_pageframe(1));
+ if (pcb->fpu_state == NULL) {
+ return -1;
+ }
+
+ /*
+ * Setup x87 FPU control word and SSE MXCSR bits
+ * as per the sysv ABI
+ */
+ __ASMV("fldcw %0\n"
+ "ldmxcsr %1"
+ :: "m" (FPU_FCW),
+ "m" (SSE_MXCSR) : "memory");
+
+ amd64_fxsave(pcb->fpu_state);
+ return 0;
+}
+
+int
+processor_free_pcb(struct proc *proc)
+{
+ struct pcb *pcb = &proc->pcb;
+
+ if (pcb->fpu_state == NULL) {
+ return -1;
+ }
+
+ vm_free_pageframe(VIRT_TO_PHYS(pcb->fpu_state), 1);
+ return 0;
+}
+
+void
+processor_switch_to(struct proc *old_td, struct proc *new_td)
+{
+ struct pcb *old_pcb = (old_td != NULL) ? &old_td->pcb : NULL;
+ struct pcb *new_pcb = &new_td->pcb;
+
+ if (old_pcb != NULL) {
+ amd64_fxsave(old_pcb->fpu_state);
+ }
+ amd64_fxrstor(new_pcb->fpu_state);
+}
+
void
processor_init(void)
{
/* Indicates what doesn't need to be init anymore */
static uint8_t init_flags = 0;
+ static uint64_t reg_tmp;
struct cpu_info *cur_cpu;
/* Create our cpu_info structure */
@@ -159,6 +224,21 @@ processor_init(void)
/* Set %GS to cpu_info */
amd64_write_gs_base((uintptr_t)cur_cpu);
+ if (is_sse_supported()) {
+ /* Enable SSE/SSE2 */
+ reg_tmp = amd64_read_cr0();
+ reg_tmp &= ~(__BIT(2));
+ reg_tmp |= __BIT(1);
+ amd64_write_cr0(reg_tmp);
+
+ /* Enable FXSAVE/FXRSTOR */
+ reg_tmp = amd64_read_cr4();
+ reg_tmp |= 3 << 9;
+ amd64_write_cr4(reg_tmp);
+ } else {
+ panic("SSE/SSE2 not supported!\n");
+ }
+
CPU_INFO_LOCK(cur_cpu);
init_tss(cur_cpu);
diff --git a/sys/arch/amd64/amd64/pmap.c b/sys/arch/amd64/amd64/pmap.c
index 2760532..54d4ca3 100644
--- a/sys/arch/amd64/amd64/pmap.c
+++ b/sys/arch/amd64/amd64/pmap.c
@@ -228,6 +228,19 @@ pmap_switch_vas(struct vm_ctx *ctx, struct vas vas)
: "memory");
}
+/*
+ * TODO: During the mapping of a virtual address, a level
+ * may be allocated. This function does not handle the
+ * freeing of allocated levels. We should keep track
+ * of levels allocated and free them here.
+ */
+int
+pmap_free_vas(struct vm_ctx *ctx, struct vas vas)
+{
+ vm_free_pageframe(vas.top_level, 1);
+ return 0;
+}
+
struct vas
pmap_read_vas(void)
{
diff --git a/sys/arch/amd64/amd64/spectre.c b/sys/arch/amd64/amd64/spectre.c
index 1247607..05aa557 100644
--- a/sys/arch/amd64/amd64/spectre.c
+++ b/sys/arch/amd64/amd64/spectre.c
@@ -62,13 +62,18 @@ __weak int
try_spectre_mitigate(void)
{
uint64_t tmp;
+ static bool should_log = true;
if (!__can_mitigate_spectre()) {
KINFO("IBRS not supported; spectre mitigation NOT enabled\n");
return EXIT_FAILURE;
}
- KINFO("IBRS supported; spectre mitigation enabled\n");
+ /* This is called per processor, only log once */
+ if (should_log) {
+ KINFO("IBRS supported; spectre mitigation enabled\n");
+ should_log = false;
+ }
tmp = rdmsr(IA32_SPEC_CTL);
tmp |= __BIT(0); /* IBRS */
diff --git a/sys/arch/amd64/amd64/syscall.S b/sys/arch/amd64/amd64/syscall.S
new file mode 100644
index 0000000..fe70523
--- /dev/null
+++ b/sys/arch/amd64/amd64/syscall.S
@@ -0,0 +1,43 @@
+/*
+ * 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/cdefs.h>
+ #include <machine/frameasm.h>
+
+__KERNEL_META "$Hyra$: syscall.S, Ian Marco Moffett, \
+ Syscall ISR code"
+
+.text
+.globl syscall_isr
+syscall_isr:
+ push_trapframe $0
+ mov %rsp, %rdi
+ call __syscall
+ pop_trapframe
+ iretq
diff --git a/sys/arch/amd64/amd64/syscall.c b/sys/arch/amd64/amd64/syscall.c
new file mode 100644
index 0000000..68235d5
--- /dev/null
+++ b/sys/arch/amd64/amd64/syscall.c
@@ -0,0 +1,49 @@
+/*
+ * 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>
+
+void
+__syscall(struct trapframe *tf)
+{
+ struct syscall_args args = {
+ .code = tf->rax,
+ .arg0 = tf->rdi,
+ .arg1 = tf->rsi,
+ .arg2 = tf->rcx,
+ .arg3 = tf->r8,
+ .arg4 = tf->r9,
+ .sp = tf->rsp
+ };
+
+ if (args.code < __MAX_SYSCALLS && args.code > 0) {
+ args.code -= 1;
+ tf->rax = g_syscall_table[args.code](&args);
+ }
+}
diff --git a/sys/arch/amd64/amd64/trap.S b/sys/arch/amd64/amd64/trap.S
index 5a77955..66dd2a9 100644
--- a/sys/arch/amd64/amd64/trap.S
+++ b/sys/arch/amd64/amd64/trap.S
@@ -144,3 +144,13 @@ nmi:
/* TODO */
cli
hlt
+
+.globl ss_fault
+ss_fault:
+ push_trapframe_ec $TRAP_SS
+
+ handle_trap
+
+ /* TODO */
+ cli
+ hlt
diff --git a/sys/arch/amd64/amd64/trap.c b/sys/arch/amd64/amd64/trap.c
index b73048d..f1e58f1 100644
--- a/sys/arch/amd64/amd64/trap.c
+++ b/sys/arch/amd64/amd64/trap.c
@@ -44,7 +44,8 @@ static const char *trap_type[] = {
[TRAP_SEGNP] = "segment not present",
[TRAP_PROTFLT] = "general protection",
[TRAP_PAGEFLT] = "page fault",
- [TRAP_NMI] = "non-maskable interrupt"
+ [TRAP_NMI] = "non-maskable interrupt",
+ [TRAP_SS] = "stack-segment fault"
};
static const int TRAP_COUNT = __ARRAY_COUNT(trap_type);
@@ -54,12 +55,17 @@ dbg_errcode(struct trapframe *tf)
{
uint64_t ec = tf->error_code;
- if (tf->trapno == TRAP_PAGEFLT) {
+ switch (tf->trapno) {
+ case TRAP_PAGEFLT:
kprintf("bits (pwui): %c%c%c%c\n",
__TEST(ec, __BIT(0)) ? 'p' : '-',
__TEST(ec, __BIT(1)) ? 'w' : '-',
__TEST(ec, __BIT(2)) ? 'u' : '-',
__TEST(ec, __BIT(4)) ? 'i' : '-');
+ break;
+ case TRAP_SS:
+ kprintf("ss: 0x%x\n", ec);
+ break;
}
}
@@ -115,5 +121,5 @@ trap_handler(struct trapframe *tf)
}
regdump(tf);
- panic("Caught pre-sched exception\n");
+ panic("Halted\n");
}
diff --git a/sys/fs/initramfs.c b/sys/fs/initramfs.c
index c0eaa0c..6da0929 100644
--- a/sys/fs/initramfs.c
+++ b/sys/fs/initramfs.c
@@ -103,7 +103,7 @@ vop_vget(struct vnode *parent, const char *name, struct vnode **vp)
return -ENOENT;
}
- if (hdr->type != TAR_TYPEFLAG_DIR) {
+ if (hdr->type == TAR_TYPEFLAG_DIR) {
vtype = VDIR;
}
@@ -149,8 +149,8 @@ static char *
get_module(const char *path, uint64_t *size) {
for (uint64_t i = 0; i < mod_req.response->module_count; ++i) {
if (strcmp(mod_req.response->modules[i]->path, path) == 0) {
- *size = mod_req.response->modules[i]->size;
- return mod_req.response->modules[i]->address;
+ *size = mod_req.response->modules[i]->size;
+ return mod_req.response->modules[i]->address;
}
}
@@ -173,6 +173,7 @@ static int
initramfs_init(struct fs_info *info)
{
initramfs = get_module("/boot/initramfs.tar", &initramfs_size);
+ info->caps = FSCAP_FULLPATH;
if (initramfs == NULL) {
panic("Failed to load initramfs\n");
@@ -221,7 +222,7 @@ initramfs_open(const char *path)
}
hdr = initramfs_from_path((void *)initramfs, path);
- return hdr_to_contents(hdr);
+ return (hdr == NULL) ? NULL : hdr_to_contents(hdr);
}
struct vfsops g_initramfs_ops = {
diff --git a/sys/include/arch/amd64/cpu.h b/sys/include/arch/amd64/cpu.h
index e2ed851..ee9bad7 100644
--- a/sys/include/arch/amd64/cpu.h
+++ b/sys/include/arch/amd64/cpu.h
@@ -110,6 +110,46 @@ amd64_read_gs_base(void)
return rdmsr(IA32_KERNEL_GS_BASE);
}
+static inline uint64_t
+amd64_read_cr0(void)
+{
+ uint64_t cr0;
+ __ASMV("mov %%cr0, %0" : "=r" (cr0) :: "memory");
+ return cr0;
+}
+
+static inline void
+amd64_write_cr0(uint64_t val)
+{
+ __ASMV("mov %0, %%cr0" :: "r" (val) : "memory");
+}
+
+static inline uint64_t
+amd64_read_cr4(void)
+{
+ uint64_t cr4;
+ __ASMV("mov %%cr4, %0" : "=r" (cr4) :: "memory");
+ return cr4;
+}
+
+static inline void
+amd64_write_cr4(uint64_t val)
+{
+ __ASMV("mov %0, %%cr4" :: "r" (val) : "memory");
+}
+
+static inline void
+amd64_fxsave(void *area)
+{
+ __ASMV("fxsave (%0)" :: "r" (area) : "memory");
+}
+
+static inline void
+amd64_fxrstor(void *area)
+{
+ __ASMV("fxrstor (%0)" :: "r" (area) : "memory");
+}
+
struct cpu_info *amd64_this_cpu(void);
#endif /* !_AMD64_CPU_H_ */
diff --git a/sys/include/arch/amd64/frame.h b/sys/include/arch/amd64/frame.h
index a2e1a05..298a836 100644
--- a/sys/include/arch/amd64/frame.h
+++ b/sys/include/arch/amd64/frame.h
@@ -67,5 +67,12 @@ struct trapframe {
(FRAME)->rsp = SP; \
(FRAME)->ss = 0x10; \
+#define init_frame_user(FRAME, IP, SP) \
+ (FRAME)->rip = IP; \
+ (FRAME)->cs = 0x18 | 3; \
+ (FRAME)->rflags = 0x202; \
+ (FRAME)->rsp = SP; \
+ (FRAME)->ss = 0x20 | 3; \
+
#endif /* !defined(__ASSEMBLER__) */
#endif /* !_AMD64_FRAME_H_ */
diff --git a/sys/include/arch/amd64/pcb.h b/sys/include/arch/amd64/pcb.h
new file mode 100644
index 0000000..0e0aab8
--- /dev/null
+++ b/sys/include/arch/amd64/pcb.h
@@ -0,0 +1,39 @@
+/*
+ * 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.
+ */
+
+#ifndef _AMD64_PCB_H_
+#define _AMD64_PCB_H_
+
+#include <sys/types.h>
+
+struct pcb {
+ uint8_t *fpu_state;
+};
+
+#endif /* !_AMD64_PCB_H_ */
diff --git a/sys/include/arch/amd64/trap.h b/sys/include/arch/amd64/trap.h
index 1019999..c75fa28 100644
--- a/sys/include/arch/amd64/trap.h
+++ b/sys/include/arch/amd64/trap.h
@@ -47,6 +47,7 @@
#define TRAP_PROTFLT 9 /* General protection */
#define TRAP_PAGEFLT 10 /* Page fault */
#define TRAP_NMI 11 /* Non-maskable interrupt */
+#define TRAP_SS 12 /* Stack-segment fault */
/* Trap is coming from user mode */
#define TRAP_USER 0x100
@@ -65,6 +66,7 @@ void segnp(void *sf);
void general_prot(void *sf);
void page_fault(void *sf);
void nmi(void *sf);
+void ss_fault(void *sf);
void trap_handler(struct trapframe *tf);
#else
.macro handle_trap
diff --git a/sys/include/sys/filedesc.h b/sys/include/sys/filedesc.h
new file mode 100644
index 0000000..ef74fb1
--- /dev/null
+++ b/sys/include/sys/filedesc.h
@@ -0,0 +1,53 @@
+/*
+ * 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.
+ */
+
+#ifndef _SYS_FILEDESC_H_
+#define _SYS_FILEDESC_H_
+
+#include <sys/vnode.h>
+#include <sys/spinlock.h>
+#include <sys/types.h>
+
+struct proc;
+
+struct filedesc {
+ int fdno;
+ off_t offset;
+ bool is_dir;
+ struct vnode *vnode;
+ struct spinlock lock;
+};
+
+#if defined(_KERNEL)
+struct filedesc *fd_alloc(struct proc *td);
+struct filedesc *fd_from_fdnum(const struct proc *td, int fdno);
+void fd_close_fdnum(struct proc *td, int fdno);
+#endif
+
+#endif
diff --git a/sys/include/sys/loader.h b/sys/include/sys/loader.h
index c1aa426..74a325c 100644
--- a/sys/include/sys/loader.h
+++ b/sys/include/sys/loader.h
@@ -33,17 +33,22 @@
#include <sys/types.h>
#include <vm/pmap.h>
-#define AT_NULL 0
-#define AT_IGNORE 1
-#define AT_EXECFD 2
-#define AT_PHDR 3
-#define AT_PHENT 4
-#define AT_PHNUM 5
-#define AT_PAGESZ 6
-#define AT_BASE 7
-#define AT_FLAGS 8
-#define AT_ENTRY 9
-#define AT_SECURE 10
+/* DANGER!: DO NOT CHANGE THESE DEFINES */
+#define AT_NULL 0
+#define AT_ENTRY 1
+#define AT_PHDR 2
+#define AT_PHENT 3
+#define AT_PHNUM 4
+#define AT_EXECPATH 5
+#define AT_SECURE 6
+#define AT_RANDOM 7
+#define AT_EXECFN 8
+
+#define STACK_PUSH(ptr, val) *(--ptr) = val
+#define AUXVAL(ptr, tag, val) __extension__ ({ \
+ STACK_PUSH(ptr, val); \
+ STACK_PUSH(ptr, tag); \
+});
/* Auxiliary Vector */
struct auxval {
diff --git a/sys/include/sys/machdep.h b/sys/include/sys/machdep.h
index 713b7db..0c06374 100644
--- a/sys/include/sys/machdep.h
+++ b/sys/include/sys/machdep.h
@@ -32,11 +32,15 @@
#include <sys/types.h>
#include <sys/cdefs.h>
+#include <sys/proc.h>
#if defined(_KERNEL)
#define MAXCPUS 32
+int processor_init_pcb(struct proc *proc);
+int processor_free_pcb(struct proc *proc);
+void processor_switch_to(struct proc *old_td, struct proc *new_td);
void processor_init(void);
void processor_halt(void);
void intr_mask(void);
diff --git a/sys/include/sys/mount.h b/sys/include/sys/mount.h
index 3ac7ec7..209fa3e 100644
--- a/sys/include/sys/mount.h
+++ b/sys/include/sys/mount.h
@@ -33,6 +33,7 @@
#include <sys/types.h>
#include <sys/queue.h>
#include <sys/vnode.h>
+#include <sys/cdefs.h>
#define FS_NAME_MAX 16 /* Max length of FS type name including nul */
@@ -54,9 +55,15 @@ struct fs_info {
char name[FS_NAME_MAX]; /* Filesystem type name */
struct vfsops *vfsops; /* Filesystem operations */
struct mount *mp_root;
+ uint16_t caps;
};
/*
+ * Filesystem capabilities
+ */
+#define FSCAP_FULLPATH __BIT(0) /* Requires full path per lookup */
+
+/*
* Mount flags
*/
#define MNT_RDONLY 0x00000001
diff --git a/sys/include/sys/proc.h b/sys/include/sys/proc.h
index f45e4c6..c6046d7 100644
--- a/sys/include/sys/proc.h
+++ b/sys/include/sys/proc.h
@@ -32,8 +32,13 @@
#include <sys/types.h>
#include <sys/queue.h>
+#include <sys/filedesc.h>
#include <machine/cpu.h>
#include <machine/frame.h>
+#include <machine/pcb.h>
+#include <vm/vm.h>
+
+#define PROC_MAX_FDS 256
/*
* A task running on the CPU e.g., a process or
@@ -43,6 +48,11 @@ struct proc {
pid_t pid;
struct cpu_info *cpu;
struct trapframe *tf;
+ struct pcb pcb;
+ struct vas addrsp;
+ uintptr_t stack_base;
+ uint8_t is_user;
+ struct filedesc *fds[PROC_MAX_FDS];
TAILQ_ENTRY(proc) link;
};
diff --git a/sys/include/sys/sched.h b/sys/include/sys/sched.h
index d803df0..1fa947e 100644
--- a/sys/include/sys/sched.h
+++ b/sys/include/sys/sched.h
@@ -37,7 +37,9 @@
#include <machine/cpu.h>
#include <machine/frame.h>
+struct proc *this_td(void);
void sched_init(void);
+void sched_exit(void);
void sched_context_switch(struct trapframe *tf);
__noreturn
diff --git a/sys/include/sys/syscall.h b/sys/include/sys/syscall.h
new file mode 100644
index 0000000..66dc5f3
--- /dev/null
+++ b/sys/include/sys/syscall.h
@@ -0,0 +1,56 @@
+/*
+ * 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.
+ */
+
+#ifndef _SYS_SYSCALL_H_
+#define _SYS_SYSCALL_H_
+
+#include <sys/types.h>
+#if defined(_KERNEL)
+#include <machine/frame.h>
+#endif
+
+/* Do not reorder */
+enum {
+ SYS_exit = 1,
+ __MAX_SYSCALLS
+};
+
+struct syscall_args {
+ uint64_t code;
+ uint64_t arg0, arg1, arg2, arg3, arg4;
+ uint64_t ip;
+ uint64_t sp;
+};
+
+#if defined(_KERNEL)
+extern uint64_t(*g_syscall_table[__MAX_SYSCALLS])(struct syscall_args *args);
+void __syscall(struct trapframe *tf);
+#endif
+
+#endif
diff --git a/sys/include/sys/vfs.h b/sys/include/sys/vfs.h
index a684c3f..c1bef53 100644
--- a/sys/include/sys/vfs.h
+++ b/sys/include/sys/vfs.h
@@ -42,7 +42,7 @@ void vfs_init(void);
struct fs_info *vfs_byname(const char *name);
int vfs_vget(struct vnode *parent, const char *name, struct vnode **vp);
-struct vnode *vfs_path_to_node(const char *path);
+int vfs_path_to_node(const char *path, struct vnode **vp);
char *vfs_get_fname_at(const char *path, size_t idx);
int vfs_rootname(const char *path, char **new_path);
bool vfs_is_valid_path(const char *path);
diff --git a/sys/include/sys/vnode.h b/sys/include/sys/vnode.h
index f584356..545e38f 100644
--- a/sys/include/sys/vnode.h
+++ b/sys/include/sys/vnode.h
@@ -47,6 +47,7 @@ struct vnode {
struct mount *mp; /* Ptr to vfs vnode is in */
struct vops *vops;
struct vnode *parent;
+ struct fs_info *fs; /* Filesystem this vnode belongs to, can be NULL */
void *data; /* Filesystem specific data */
};
diff --git a/sys/include/vm/pmap.h b/sys/include/vm/pmap.h
index ebabd32..3380199 100644
--- a/sys/include/vm/pmap.h
+++ b/sys/include/vm/pmap.h
@@ -90,6 +90,12 @@ struct vas pmap_read_vas(void);
int pmap_map(struct vm_ctx *, struct vas, vaddr_t, paddr_t, vm_prot_t);
/*
+ * Get rid of a virtual address space and free
+ * resources.
+ */
+int pmap_free_vas(struct vm_ctx *, struct vas);
+
+/*
* Unmap a page.
*/
int pmap_unmap(struct vm_ctx *, struct vas, vaddr_t);
diff --git a/sys/include/vm/vm.h b/sys/include/vm/vm.h
index 2a24d76..48e1b8f 100644
--- a/sys/include/vm/vm.h
+++ b/sys/include/vm/vm.h
@@ -61,5 +61,6 @@ vm_get_page_size(void)
void vm_init(void);
struct vm_ctx *vm_get_ctx(void);
+struct vas vm_get_kvas(void);
#endif /* !_VM_H_ */
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;
}
diff --git a/sys/vm/vm_init.c b/sys/vm/vm_init.c
index f15cb0b..6096059 100644
--- a/sys/vm/vm_init.c
+++ b/sys/vm/vm_init.c
@@ -59,6 +59,15 @@ vm_get_ctx(void)
return &bsp_vm_ctx;
}
+/*
+ * Return the kernel VAS.
+ */
+struct vas
+vm_get_kvas(void)
+{
+ return kernel_vas;
+}
+
void
vm_init(void)
{
diff --git a/sys/vm/vm_map.c b/sys/vm/vm_map.c
index 0d27738..f65bad2 100644
--- a/sys/vm/vm_map.c
+++ b/sys/vm/vm_map.c
@@ -78,12 +78,13 @@ int
vm_map_create(struct vas vas, vaddr_t va, paddr_t pa, vm_prot_t prot, size_t bytes)
{
size_t granule = vm_get_page_size();
+ size_t misalign = va & (granule - 1);
int s;
struct vm_ctx *ctx = vm_get_ctx();
/* We want bytes to be aligned by the granule */
- bytes = __ALIGN_UP(bytes, granule);
+ bytes = __ALIGN_UP(bytes + misalign, granule);
/* Align VA/PA by granule */
va = __ALIGN_DOWN(va, granule);
@@ -115,10 +116,11 @@ vm_map_destroy(struct vas vas, vaddr_t va, size_t bytes)
{
struct vm_ctx *ctx = vm_get_ctx();
size_t granule = vm_get_page_size();
+ size_t misalign = va & (granule - 1);
int s;
/* We want bytes to be aligned by the granule */
- bytes = __ALIGN_UP(bytes, granule);
+ bytes = __ALIGN_UP(bytes + misalign, granule);
/* Align VA by granule */
va = __ALIGN_DOWN(va, granule);