Pull the latest asan

.github/workflows/main.yml

@@ -52,6 +52,7 @@ jobs:
           export ONYX_ARCH=x86_64
           ./scripts/setup_build.sh
           RUN_CLANG_TIDY=0 make -j $(nproc) iso
+          ./scripts/create_disk_image.sh disk-image.img --bootable efi
 
       - name: Upload a Build Artifact(Onyx.iso)
         uses: actions/upload-artifact@v2.1.4
@@ -65,11 +66,11 @@ jobs:
           name: vmonyx
           path: kernel/vmonyx
 
-      - name: Upload a Build Artifact(sysroot)
+      - name: Upload a Build Artifact(Disk image)
         uses: actions/upload-artifact@v2.1.4
         with:
-          name: Sysroot
-          path: sysroot/
+          name: Disk-image.img
+          path: disk-image.img
 
   build-llvm:
     # The type of runner that the job will run on
@@ -113,6 +114,7 @@ jobs:
           export ONYX_ARCH=x86_64
           ./scripts/setup_build.sh
           RUN_CLANG_TIDY=0 make -j $(nproc) iso
+          ./scripts/create_disk_image.sh disk-image-llvm.img --bootable efi
 
       - name: Upload a Build Artifact(Onyx.iso)
         uses: actions/upload-artifact@v2.1.4
@@ -126,11 +128,11 @@ jobs:
           name: vmonyx-llvm
           path: kernel/vmonyx
 
-      - name: Upload a Build Artifact(sysroot-llvm)
+      - name: Upload a Build Artifact(disk-image-llvm)
         uses: actions/upload-artifact@v2.1.4
         with:
-          name: Sysroot-llvm
-          path: sysroot/
+          name: Disk-image-llvm.img
+          path: disk-image-llvm.img
 
   # Builds a minimal sysroot, with headers and the libc. Supposed to be used when building a toolchain.
   # In this case, we limit ourselves to building a stage 1 GNU toolchain since LLVM is far slower to build,

kernel/arch/x86_64/mmu.cpp

@@ -730,12 +730,11 @@ bool paging_write_protect(void *addr, struct mm_address_space *mm)
 
 int is_invalid_arch_range(void *address, size_t pages)
 {
-	for(uintptr_t addr = (uintptr_t) address, i = 0; i < pages; ++i, addr += PAGE_SIZE)
-	{
-		// Non canonical
-		if(addr > arch_low_half_max && addr < VM_HIGHER_HALF)
-			return -1;
-	}
+	unsigned long addr = (unsigned long) address;
+	auto limit = addr + (pages << PAGE_SHIFT);
+
+	if (addr <= arch_low_half_max && limit >= VM_HIGHER_HALF)
+		return -1;
 	return 0;
 }
 

kernel/drivers/ahci/ahci.cpp

@@ -112,6 +112,12 @@ irqstatus_t ahci_irq(struct irq_context *ctx, void *cookie)
 
 		if(ports & (1U << i))
 		{
+			if(!port->port)
+			{
+				panic("what? panic at the disco #%u", i);
+				return IRQ_UNHANDLED;
+			}
+
 			uint32_t port_is = port->port->interrupt_status;
 			port->port->interrupt_status = port_is;
 			dev->hba->interrupt_status = (1U << i);
@@ -901,6 +907,9 @@ void ahci_init_port(struct ahci_port *ahci_port)
 	struct ahci_device *device = ahci_port->dev;
 	ahci_hba_memory_regs_t *hba = device->hba;
 
+	// ACK old interrupts that might've gotten stuck
+	uint32_t port_is = port->interrupt_status;
+	port->interrupt_status = port_is;
 	/* Enable interrupts */
 	ahci_enable_interrupts_for_port(port);
 
@@ -945,7 +954,7 @@ void ahci_init_port(struct ahci_port *ahci_port)
 
 	port->pxcmd = port->pxcmd | AHCI_PORT_CMD_START;
 
-	ahci_do_identify(ahci_port);
+	//ahci_do_identify(ahci_port);
 }
 
 int ahci_initialize(struct ahci_device *device)
@@ -955,9 +964,6 @@ int ahci_initialize(struct ahci_device *device)
 	/* Firstly, set the AE bit on the GHC register to indicate we're AHCI aware */
 	hba->ghc = hba->ghc | AHCI_GHC_AHCI_ENABLE;
 
-	/* Now, enable interrupts in the HBA */
-	hba->ghc = hba->ghc | AHCI_GHC_INTERRUPTS_ENABLE;
-
 	int nr_ports = AHCI_CAP_NR_PORTS(hba->host_cap);
 	if(nr_ports == 0)	nr_ports = 1;
 
@@ -1015,13 +1021,16 @@ int ahci_initialize(struct ahci_device *device)
 			device->ports[i].port_nr = i; 
 			device->ports[i].port = &hba->ports[i];
 			device->ports[i].dev = device;
+			device->ports[i].bdev = dev;
 
 			ahci_init_port(&device->ports[i]);
-
-			blkdev_init(dev);
 		}
 	}
 
+	hba->interrupt_status = hba->interrupt_status;
+	/* Now, enable interrupts in the HBA */
+	hba->ghc = hba->ghc | AHCI_GHC_INTERRUPTS_ENABLE;
+
 	return 0;
 }
 
@@ -1036,6 +1045,7 @@ int ahci_probe(struct device *dev)
 {
 	int status = 0;
 	int irq = -1;
+	int nr_ports;
 	pci::pci_device *ahci_dev = (pci::pci_device *) dev;
 
 	if(ahci_dev->enable_device() < 0)
@@ -1063,7 +1073,15 @@ int ahci_probe(struct device *dev)
 		status = -1;
 		goto ret;
 	}
-
+
+	/* Initialize AHCI */
+	if(ahci_initialize(device) < 0)
+	{
+		MPRINTF("Failed to initialize the AHCI controller\n");
+		status = -1;
+		goto ret;
+	}
+
 	if(ahci_dev->enable_msi(ahci_irq, device))
 	{
 		/* If we couldn't enable MSI, use normal I/O APIC pins */
@@ -1075,12 +1093,19 @@ int ahci_probe(struct device *dev)
 		assert(install_irq(irq, ahci_irq, (struct device *) ahci_dev,
 			IRQ_FLAG_REGULAR, device) == 0);
 	}
-	/* Initialize AHCI */
-	if(ahci_initialize(device) < 0)
+
+	nr_ports = AHCI_CAP_NR_PORTS(hba->host_cap);
+	if(nr_ports == 0)	nr_ports = 1;
+
+	// For every port in the controller, see if we have initialised it. If so, do identify and blkdev_init
+	for (int i = 0; i < nr_ports; i++)
 	{
-		MPRINTF("Failed to initialize the AHCI controller\n");
-		status = -1;
-		goto ret;
+		if (!device->ports[i].bdev)
+			continue;
+
+		ahci_do_identify(&device->ports[i]);
+
+		blkdev_init(device->ports[i].bdev);
 	}
 
 	ahci_probe_ports(count_bits<uint32_t>(hba->ports_implemented), hba);

kernel/drivers/ahci/include/ahci.h

@@ -240,6 +240,7 @@ struct ahci_port
 	struct wait_queue list_wq;
 	unsigned char identify[512];
 	uint32_t issued;
+	blockdev *bdev;
 };
 
 struct ahci_device

kernel/drivers/ata/ata.cpp

@@ -434,6 +434,11 @@ int ata_probe(struct device *d)
 {
 	pci::pci_device *device = (pci::pci_device *) d;
 
+	// If this is a SATA controller with a valid BAR5 (AHCI HBA BAR), skip this
+	// and defer to AHCI.
+	if (device->sub_class() == 6 && device->get_bar(5).has_value())
+		return -1;
+
 	unique_ptr<ide_dev> dev = make_unique<ide_dev>(device);
 	if(!dev)
 		return -ENOMEM;

kernel/drivers/pci/pci.cpp

@@ -413,6 +413,9 @@ void __pci_bus_register(struct driver *driver, struct bus *bus)
 			dev->addr().segment, dev->addr().bus, dev->addr().device, dev->addr().function,
 			dev->did(), dev->vid());
 #endif
+		// Bound, skip.
+		if(dev->driver_)
+			continue;
 
 		if((id = pci_driver_supports_device(driver, dev)))
 		{

kernel/include/onyx/x86/include/platform/vm_layout.h

@@ -15,7 +15,7 @@
 
 typedef enum
 {
-	arch_low_half_min  	= 0x400000,
+	arch_low_half_min  	= 0x200000,
 	arch_brk_base		= 0x80000000,
 	arch_mmap_base		= 0x0000550000000000,
 	arch_low_half_max  	= 0x00007fffffffffff,

kernel/kernel/fs/partition.cpp

@@ -237,7 +237,6 @@ int partition_setup_disk_gpt(struct blockdev *dev)
 			goto out;
 		}
 
-		//printk("registered! lba %lu\n", e->first_lba);
 		nr_parts++;
 	}
 

kernel/kernel/init.cpp

@@ -86,45 +86,6 @@ void *tramp = NULL;
 void *phys_fb = NULL;
 uintptr_t address = 0;
 
-extern void libc_late_init();
-
-char *kernel_arguments[200];
-int kernel_argc = 0;
-
-void kernel_parse_command_line(char *cmd)
-{
-	char *original_string = cmd;
-	while(*cmd)
-	{
-		if(*cmd == '-') /* Found an argument */
-		{
-			char *token = strchr(cmd, ' ');
-			if(!token)
-				token = original_string + strlen(original_string);
-			size_t size_token = (size_t)(token - cmd);
-			char *new_string = (char *) malloc(size_token + 1);
-			if(!new_string)
-			{
-				ERROR("kernel", "failed parsing: out of memory\n");
-			}
-
-			memset(new_string, 0, size_token + 1);
-			memcpy(new_string, cmd, size_token);
-
-			if(kernel_argc == 200)
-			{
-				panic("kernel: too many arguments passed to the kernel");
-			}
-
-			kernel_arguments[kernel_argc] = new_string;
-			kernel_argc++;
-			cmd += size_token - 1;
-		}
-
-		cmd++;
-	}
-}
-
 void set_initrd_address(void *initrd_address)
 {
 	initrd_addr = initrd_address;
@@ -332,7 +293,7 @@ void kernel_multitasking(void *arg)
 	vm_sysfs_init();
 
 	/* Pass the root partition to init */
-	const char *args[] = {(char *) "", NULL};
+	const char *args[] = {(char *) "", "/dev/sda2", NULL};
 	const char *envp[] = {"PATH=/bin:/usr/bin:/sbin:", "TERM=linux", "LANG=C", "PWD=/", NULL};
 
 	if(find_and_exec_init(args, envp) < 0)

kernel/kernel/process.cpp

@@ -837,7 +837,8 @@ void process_exit(unsigned int exit_code)
 		printk("Panic: %s exited with exit code %u!\n",
 			current->cmd_line.c_str(), exit_code);
 		irq_enable();
-		for(;;);
+		for(;;)
+			sched_sleep_ms(10000);
 	}
 
 	process_kill_other_threads();

scripts/build_grub_efi_image.sh

@@ -0,0 +1,12 @@
+#!/bin/sh
+# Copyright (c) 2021 Pedro Falcato
+# This file is part of Onyx, and is released under the terms of the MIT License
+# check LICENSE at the root directory for more information
+#
+# SPDX-License-Identifier: MIT
+#
+
+grub-mkimage -p /boot/grub -O "${ONYX_ARCH}-efi" -c scripts/grub/grub-early.cfg \
+-o bootx64.efi at_keyboard boot chain configfile fat ext2 multiboot multiboot2 \
+linux ls part_gpt reboot serial efi_gop efi_uga echo elf ata ahci gfxterm all_video xzio \
+font multiboot

scripts/ci/install_ubuntu_deps.sh

@@ -1,7 +1,8 @@
 #!/bin/sh
 set -e
 
-sudo apt-get update && sudo apt-get install libmpc-dev libgmp3-dev bison flex libmpfr-dev zstd ninja-build clang lld
+sudo apt-get update && sudo apt-get install libmpc-dev libgmp3-dev bison flex libmpfr-dev zstd ninja-build clang lld \
+parted mtools
 mkdir gn_bin/
 cd gn_bin
 wget -q https://chrome-infra-packages.appspot.com/dl/gn/gn/linux-amd64/+/latest -O gn.zip

scripts/create_adaptive_disk_image.py

@@ -0,0 +1,30 @@
+#!/bin/python3
+# Copyright (c) 2021 Pedro Falcato
+# This file is part of Onyx, and is released under the terms of the MIT License
+# check LICENSE at the root directory for more information
+#
+# SPDX-License-Identifier: MIT
+#
+
+import sys
+import subprocess
+
+def main():
+    if len(sys.argv) < 3:
+        print("create_adaptive_disk_image.py: Bad usage")
+        exit(1)
+
+    # Our argument is whatever du(1) calculated, in 1024-byte units
+    nr_1024_blocks = int(sys.argv[1])
+
+    nr_1024_blocks = nr_1024_blocks * 1024
+    # Have the final partition image be 20% larger than the total size
+    # Note that it might be a decent idea to have this be adjustable by the user
+    # or maybe we get could a better estimate.
+    nr_1024_blocks = int(nr_1024_blocks * 1.20)
+
+    return subprocess.run(["fallocate", "-l", str(nr_1024_blocks), sys.argv[2]]).returncode
+
+if __name__ == "__main__":
+    main()
+