Hardware fixes & USB updates

crates/display-server/src/main.rs

@@ -29,7 +29,7 @@ fn handle_incoming(_msg: ulib::sys::Message, _buf: &[u8], resp_socket: FileDesc)
     // TODO: proper listen + connect sockets
     // (this just broadcasts a response to all listeners and hopes that there aren't race conditions)
 
-    let buffer = init_buffer();
+    let buffer = init_buffer(640, 480);
 
     let fds = [buffer.fd, buffer.present_sem_fd];
     let handle = unsafe { proto::BufferHandle::new(buffer.mapped, &fds) };
@@ -55,10 +55,9 @@ fn handle_incoming(_msg: ulib::sys::Message, _buf: &[u8], resp_socket: FileDesc)
     handle
 }
 
-fn init_buffer() -> BufferInfo {
+fn init_buffer(width: usize, height: usize) -> BufferInfo {
     println!("init_buffer");
-    let screen_size = (640, 480);
-    let vmem_size = screen_size.0 * screen_size.1 * 4;
+    let vmem_size = width * height * 4;
 
     let header_size = size_of::<proto::BufferHeader>().next_multiple_of(4096);
     let total_size = header_size + vmem_size;
@@ -87,9 +86,9 @@ fn init_buffer() -> BufferInfo {
         server_to_client_queue: proto::EventQueue::new(),
 
         video_meta: proto::VideoMeta {
-            width: screen_size.0 as u16,
-            height: screen_size.1 as u16,
-            row_stride: (screen_size.0 as u16 * 4),
+            width: width as u16,
+            height: height as u16,
+            row_stride: (width as u16 * 4),
             bytes_per_pixel: 4,
             bit_layout: 0,
             present_ts: 0,

crates/kernel/.gitignore

@@ -0,0 +1 @@
+*.pcap

crates/kernel/examples/timer.rs

@@ -6,7 +6,7 @@ extern crate kernel;
 
 use kernel::*;
 
-fn test_irq_handler(_ctx: &mut event::context::Context) {
+fn test_irq_handler(_ctx: &mut event::context::Context, _irq: usize) {
     //Reset the timer to ping again
     device::system_timer::ARM_GENERIC_TIMERS.with_current(|timer| {
         timer.reset_timer();

crates/kernel/examples/user.rs

@@ -42,6 +42,8 @@ async fn main() {
     let user_region = user_region as *mut u8;
 
     let ttbr0 = process.get_ttbr0();
+    println!("init ttbr0={:#x}", ttbr0);
+
     let callback = || {
         println!("User ptr: {:p}", user_region);
         println!(
@@ -65,6 +67,13 @@ async fn main() {
         let end = sync::get_time();
 
         println!("Done copying user data, took {:4}µs", end - start);
+
+        unsafe {
+            crate::arch::memory::flush_range(
+                user_region.addr(),
+                user_region.addr() + INIT_CODE.len(),
+            )
+        };
     };
     unsafe { with_user_vmem(ttbr0, callback) };
 

crates/kernel/scripts/run-usb-passthrough.sh

@@ -0,0 +1,8 @@
+#!/usr/bin/env bash
+
+set -ex
+QEMU_DISPLAY=${QEMU_DISPLAY-"default"} \
+QEMU_DEVICES="-usb -device usb-host,hostbus=3,hostport=4.3" "$(dirname "$0")/run.sh"
+#QEMU_DEVICES="-usb -device usb-host,vendorid=0x1532,productid=0x025e" "$(dirname "$0")/run.sh"
+#QEMU_DEVICES="-usb -device usb-host,vendorid=0x2109,productid=0x2817" "$(dirname "$0")/run.sh"
+

crates/kernel/scripts/run-usb.sh

@@ -1,4 +1,5 @@
 #!/usr/bin/env bash
 
 set -ex
-QEMU_DISPLAY="default" QEMU_DEVICES="-usb -device usb-kbd -device usb-mouse -device usb-net,netdev=net0 -netdev user,id=net0,hostfwd=tcp::2222-:22 -object filter-dump,id=f1,netdev=net0,file=net0.pcap" "$(dirname "$0")/run.sh"
+QEMU_DISPLAY=${QEMU_DISPLAY-"default"} QEMU_DEVICES="-device usb-hub,id=hub1 -device usb-kbd -device usb-mouse" "$(dirname "$0")/run.sh"
+# QEMU_DISPLAY=${QEMU_DISPLAY-"default"} QEMU_DEVICES="-usb -device usb-kbd -device usb-mouse,pcap=usb-mouse.pcap -device usb-net,netdev=net0 -netdev user,id=net0,hostfwd=tcp::2222-:22 -object filter-dump,id=f1,netdev=net0,file=net0.pcap" "$(dirname "$0")/run.sh"

crates/kernel/src/arch/aarch64/boot.rs

@@ -83,16 +83,20 @@ kernel_entry_alt:
 
     b kernel_entry_rust_alt
 
-    // TODO: somehow this sometimes gets triggered twice on core 0?
-    // Taking exception 1 [Undefined Instruction] on CPU 0
-    // ...from EL1 to EL1
-    // ...with ESR 0x0/0x2000000
-    // ...with ELR 0x8005c
-    // ...to EL1 PC 0x80a00 PSTATE 0x3c5
 drop_to_el1:
+    // Init core ids?  TODO: probably unneeded
+    mrs x5, midr_el1
+    mrs x6, mpidr_el1
+    msr vpidr_el2, x5
+    msr vmpidr_el2, x6
+
+    // Init timers on core zero???
+    mrs x5, cnthctl_el2
+    orr x5, x5, #0x3
+    msr cnthctl_el2, x5
+    msr cntvoff_el2, xzr
 
-    mov x5, #(1 << 31)
-    // orr x5, x5, #0x38
+    ldr x5, ={HCR_EL2}
     msr hcr_el2, x5
 
     // Enable FPU
@@ -114,7 +118,7 @@ drop_to_el1:
     ldr x5, ={MAIR_EL1}
     msr MAIR_EL1, x5
 
-    mov x5, #0b0101
+    mov x5, #0x3c5 // EL1_SP1, DAIF masked
     msr SPSR_EL2, x5
 
     // Enable FPU
@@ -163,6 +167,7 @@ switch_kernel_vmem:
     mrs x3, TTBR1_EL1
     msr TTBR0_EL1, x3
 
+    isb
     dsb sy
     tlbi vmalle1is
     dsb sy
@@ -173,6 +178,7 @@ switch_kernel_vmem:
 
     msr TTBR0_EL1, x4
 
+    isb
     dsb sy
     tlbi vmalle1is
     dsb sy
@@ -187,6 +193,7 @@ switch_kernel_vmem_in_phys:
     msr TCR_EL1, x1
     msr TTBR1_EL1, x0
 
+    isb
     dsb sy
     tlbi vmalle1is
     dsb sy
@@ -198,6 +205,7 @@ switch_kernel_vmem_in_phys:
 switch_user_tcr_el1:
     msr TCR_EL1, x0
 
+    isb
     dsb sy
     tlbi vmalle1is
     dsb sy
@@ -207,17 +215,28 @@ switch_user_tcr_el1:
     STACK_SIZE_LOG2 = const STACK_SIZE_LOG2,
     TCR_EL1 = const INIT_TCR_EL1,
     TRANSLATION_ENTRY = const INIT_TRANSLATION,
+    HCR_EL2 = const (
+        (1 << 31) | // 64 bit mode in EL1
+        // orr x5, x5, #0x38 (???)
+        0
+    ),
     SCTLR_EL1 = const (
-        (1 << 11) | // enable instruction caching
+        (1 << 29) | // ???
+        (1 << 28) |
+        (1 << 23) |
+        (1 << 22) |
+        (1 << 20) |
+        (1 << 12) | // enable instruction caching
+        (1 << 11) | // exceptions as sync point?
         (1 << 4) | // enable EL0 stack pointer alignment
         (1 << 3) | // enable EL1 stack pointer alignment
         (1 << 2) | // enable data caching
         (1 << 1) | // enable alignment faults
         1           // enable EL1&0 virtual memory
     ),
     MAIR_EL1 = const (
-        (0b01000100 << 16) | // Entry 2: normal noncacheable memory
-        (0b00000000 << 8)  | // Entry 1: device memory
-        (0b11111111 << 0)    // Entry 0: normal memory
+        (0x44 << 16) | // Entry 2: normal noncacheable memory
+        (0x00 << 8)  | // Entry 1: device memory
+        (0xFF << 0)    // Entry 0: normal memory
     )
 );

crates/kernel/src/arch/aarch64/memory/mod.rs

@@ -97,42 +97,55 @@ pub unsafe fn clean_physical_buffer_for_device(va: *mut (), bytes: usize) {
     for ptr in va..(va + bytes) {
         // clean each byte
         // TODO: only invoke the cleaning once per cache line by using the cache registers to find line width
-        unsafe {
-            asm! {
-                "dc cvac, {ptr}",
-                ptr = in(reg) ptr,
-                options(readonly, nostack, preserves_flags)
-            }
-        }
+        unsafe { asm!("dc cvac, {ptr}",ptr = in(reg) ptr,options(nostack, preserves_flags)) };
     }
     // enforce memory barrier between this and subsequent memory operations
     // must be inserted at some point before the device access, and this is a reasonable point
-    unsafe {
-        asm! {
-            "dmb sy",
-            options(readonly, nostack, preserves_flags)
-        }
-    }
+    unsafe { asm!("dmb sy", options(nostack, preserves_flags)) };
 }
 pub unsafe fn invalidate_physical_buffer_for_device(va: *mut (), bytes: usize) {
     // enforce memory barrier between this and prior memory operations
     // probably needs to be inserted (?) at some point after the device work completes, and this is a reasonable point
-    unsafe {
-        asm! {
-            "dmb sy",
-            options(readonly, nostack, preserves_flags)
-        }
-    }
+    unsafe { asm!("dmb sy", options(nostack, preserves_flags)) };
     let va = va.addr();
     for ptr in va..(va + bytes) {
         // invalidate each byte
         // TODO: only invoke the invalidating once per cache line by using the cache registers to find line width
-        unsafe {
-            asm! {
-                "dc ivac, {ptr}",
-                ptr = in(reg) ptr,
-                options(readonly, nostack, preserves_flags)
-            }
-        }
+        unsafe { asm!("dc ivac, {ptr}",ptr = in(reg) ptr,options(nostack, preserves_flags)) };
+    }
+}
+
+pub unsafe fn flush_range(start: usize, end: usize) {
+    use core::arch::asm;
+
+    // might not be needed
+    unsafe { asm!("dsb sy", "isb", options(nostack, preserves_flags)) };
+
+    let ctr_el0: usize;
+    unsafe { asm!("mrs {0}, CTR_EL0", out(reg) ctr_el0, options(nomem, nostack, preserves_flags)) };
+
+    let d_cacheline = 4 << ((ctr_el0 >> 16) & 0x0F);
+    let i_cacheline = 4 << ((ctr_el0 >> 0) & 0x0F);
+
+    let start_line = (start / d_cacheline) * d_cacheline;
+    let end_line = end.next_multiple_of(d_cacheline);
+    for line in (start_line..end_line).step_by(d_cacheline) {
+        // TODO: figure out what subset of cache flushing is needed here
+        // unsafe { asm!("dc cvau, {0}", in(reg) line, options(nostack, preserves_flags)) };
+        unsafe { asm!("dc cvac, {0}", in(reg) line, options(nostack, preserves_flags)) };
+        unsafe { asm!("dc ivac, {0}", in(reg) line, options(nostack, preserves_flags)) };
+    }
+
+    unsafe { asm!("dsb ish", options(nostack, preserves_flags)) };
+
+    let start_line = (start / i_cacheline) * i_cacheline;
+    let end_line = end.next_multiple_of(i_cacheline);
+    for line in (start_line..end_line).step_by(i_cacheline) {
+        unsafe { asm!("ic ivau, {0}", in(reg) line, options(nostack, preserves_flags)) };
     }
+
+    unsafe { asm!("dsb ish", "isb", options(nostack, preserves_flags)) };
+
+    // might not be needed
+    unsafe { asm!("dsb sy", "isb", options(nostack, preserves_flags)) };
 }