capsules_system/process_printer.rs
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252
// Licensed under the Apache License, Version 2.0 or the MIT License.
// SPDX-License-Identifier: Apache-2.0 OR MIT
// Copyright Tock Contributors 2024.
//! Tools for displaying process state.
use core::fmt::Write;
use kernel::process::Process;
use kernel::process::{ProcessPrinter, ProcessPrinterContext};
use kernel::utilities::binary_write::BinaryWrite;
use kernel::utilities::binary_write::WriteToBinaryOffsetWrapper;
/// A Process Printer that displays a process as a human-readable string.
pub struct ProcessPrinterText {}
impl ProcessPrinterText {
pub fn new() -> ProcessPrinterText {
ProcessPrinterText {}
}
}
impl ProcessPrinter for ProcessPrinterText {
// `print_overview()` must be synchronous, but does not assume a synchronous
// writer or an infinite (or very large) underlying buffer in the writer. To
// do this, this implementation assumes the underlying writer _is_
// synchronous. This makes the printing code cleaner, as it does not need to
// be broken up into chunks of some length (which would need to match the
// underlying buffer length). However, not all writers are synchronous, so
// this implementation keeps track of how many bytes were sent on the last
// call, and only prints new bytes on the next call. This works by having
// the function start from the beginning each time, formats the entire
// overview message, and just drops bytes until getting back to where it
// left off on the last call.
//
// ### Assumptions
//
// This implementation makes two assumptions:
// 1. That `print_overview()` is not called in performance-critical code.
// Since each time it formats and "prints" the message starting from the
// beginning, it duplicates a fair bit of formatting work. Since this is
// for debugging, the performance impact of that shouldn't matter.
// 2. That `printer_overview()` will be called in a tight loop, and no
// process state will change between calls. That could change the length
// of the printed message, and lead to gaps or parts of the overview
// being duplicated. However, it does not make sense that the kernel
// would want to run the process while it is displaying debugging
// information about it, so this should be a safe assumption.
fn print_overview(
&self,
process: &dyn Process,
writer: &mut dyn BinaryWrite,
context: Option<ProcessPrinterContext>,
) -> Option<ProcessPrinterContext> {
let offset = context.map_or(0, |c| c.offset);
// Process statistics
let events_queued = process.pending_tasks();
let syscall_count = process.debug_syscall_count();
let dropped_upcall_count = process.debug_dropped_upcall_count();
let restart_count = process.get_restart_count();
let addresses = process.get_addresses();
let sizes = process.get_sizes();
let process_struct_memory_location = addresses.sram_end
- sizes.grant_pointers
- sizes.upcall_list
- sizes.process_control_block;
let sram_grant_size = process_struct_memory_location - addresses.sram_grant_start;
let mut bww = WriteToBinaryOffsetWrapper::new(writer);
bww.set_offset(offset);
let _ = bww.write_fmt(format_args!(
"\
𝐀𝐩𝐩: {} - [{:?}]\
\r\n Events Queued: {} Syscall Count: {} Dropped Upcall Count: {}\
\r\n Restart Count: {}\
\r\n",
process.get_process_name(),
process.get_state(),
events_queued,
syscall_count,
dropped_upcall_count,
restart_count,
));
let _ = match process.debug_syscall_last() {
Some(syscall) => bww.write_fmt(format_args!(" Last Syscall: {:?}\r\n", syscall)),
None => bww.write_str(" Last Syscall: None\r\n"),
};
let _ = match process.get_completion_code() {
Some(opt_cc) => match opt_cc {
Some(cc) => bww.write_fmt(format_args!(" Completion Code: {}\r\n", cc as isize)),
None => bww.write_str(" Completion Code: Faulted\r\n"),
},
None => bww.write_str(" Completion Code: None\r\n"),
};
let _ = bww.write_fmt(format_args!(
"\
\r\n\
\r\n ╔═══════════╤══════════════════════════════════════════╗\
\r\n ║ Address │ Region Name Used | Allocated (bytes) ║\
\r\n ╚{:#010X}═╪══════════════════════════════════════════╝\
\r\n │ Grant Ptrs {:6}\
\r\n │ Upcalls {:6}\
\r\n │ Process {:6}\
\r\n {:#010X} ┼───────────────────────────────────────────\
\r\n │ ▼ Grant {:6}\
\r\n {:#010X} ┼───────────────────────────────────────────\
\r\n │ Unused\
\r\n {:#010X} ┼───────────────────────────────────────────",
addresses.sram_end,
sizes.grant_pointers,
sizes.upcall_list,
sizes.process_control_block,
process_struct_memory_location,
sram_grant_size,
addresses.sram_grant_start,
addresses.sram_app_brk,
));
// We check to see if the underlying writer has more work to do. If it
// does, then its buffer is full and any additional writes are just
// going to be dropped. So, we skip doing more printing if there are
// bytes remaining as a slight performance optimization.
if !bww.bytes_remaining() {
match addresses.sram_heap_start {
Some(sram_heap_start) => {
let sram_heap_size = addresses.sram_app_brk - sram_heap_start;
let sram_heap_allocated = addresses.sram_grant_start - sram_heap_start;
let _ = bww.write_fmt(format_args!(
"\
\r\n │ ▲ Heap {:6} | {:6}{} S\
\r\n {:#010X} ┼─────────────────────────────────────────── R",
sram_heap_size,
sram_heap_allocated,
exceeded_check(sram_heap_size, sram_heap_allocated),
sram_heap_start,
));
}
None => {
let _ = bww.write_str(
"\
\r\n │ ▲ Heap ? | ? S\
\r\n ?????????? ┼─────────────────────────────────────────── R",
);
}
}
}
if !bww.bytes_remaining() {
match (addresses.sram_heap_start, addresses.sram_stack_top) {
(Some(sram_heap_start), Some(sram_stack_top)) => {
let sram_data_size = sram_heap_start - sram_stack_top;
let sram_data_allocated = sram_data_size;
let _ = bww.write_fmt(format_args!(
"\
\r\n │ Data {:6} | {:6} A",
sram_data_size, sram_data_allocated,
));
}
_ => {
let _ = bww.write_str(
"\
\r\n │ Data ? | ? A",
);
}
}
}
if !bww.bytes_remaining() {
match (addresses.sram_stack_top, addresses.sram_stack_bottom) {
(Some(sram_stack_top), Some(sram_stack_bottom)) => {
let sram_stack_size = sram_stack_top - sram_stack_bottom;
let sram_stack_allocated = sram_stack_top - addresses.sram_start;
let _ = bww.write_fmt(format_args!(
"\
\r\n {:#010X} ┼─────────────────────────────────────────── M\
\r\n │ ▼ Stack {:6} | {:6}{}",
sram_stack_top,
sram_stack_size,
sram_stack_allocated,
exceeded_check(sram_stack_size, sram_stack_allocated),
));
}
_ => {
let _ = bww.write_str(
"\
\r\n ?????????? ┼─────────────────────────────────────────── M\
\r\n │ ▼ Stack ? | ?",
);
}
}
}
if !bww.bytes_remaining() {
let flash_protected_size = addresses.flash_non_protected_start - addresses.flash_start;
let flash_app_size = addresses.flash_end - addresses.flash_non_protected_start;
let _ = bww.write_fmt(format_args!(
"\
\r\n {:#010X} ┼───────────────────────────────────────────\
\r\n │ Unused\
\r\n {:#010X} ┴───────────────────────────────────────────\
\r\n .....\
\r\n {:#010X} ┬─────────────────────────────────────────── F\
\r\n │ App Flash {:6} L\
\r\n {:#010X} ┼─────────────────────────────────────────── A\
\r\n │ Protected {:6} S\
\r\n {:#010X} ┴─────────────────────────────────────────── H\
\r\n",
addresses.sram_stack_bottom.unwrap_or(0),
addresses.sram_start,
addresses.flash_end,
flash_app_size,
addresses.flash_non_protected_start,
flash_protected_size,
addresses.flash_start
));
}
if bww.bytes_remaining() {
// The underlying writer is indicating there are still bytes
// remaining to be sent. That means we want to return a context so
// the caller knows to call us again and we can keep printing until
// we have displayed the entire process overview.
let new_context = ProcessPrinterContext {
offset: bww.get_index(),
};
Some(new_context)
} else {
None
}
}
}
/// If `size` is greater than `allocated` then it returns a warning string to
/// help with debugging.
fn exceeded_check(size: usize, allocated: usize) -> &'static str {
if size > allocated {
" EXCEEDED!"
} else {
" "
}
}