diff options
-rw-r--r-- | fuzz/fuzz_targets/display_does_not_panic.rs | 18 | ||||
-rw-r--r-- | src/protected_mode/display.rs | 104 | ||||
-rw-r--r-- | src/protected_mode/mod.rs | 2 | ||||
-rw-r--r-- | src/real_mode/display.rs | 104 | ||||
-rw-r--r-- | src/real_mode/mod.rs | 2 |
5 files changed, 227 insertions, 3 deletions
diff --git a/fuzz/fuzz_targets/display_does_not_panic.rs b/fuzz/fuzz_targets/display_does_not_panic.rs index 97a14b8..39f5753 100644 --- a/fuzz/fuzz_targets/display_does_not_panic.rs +++ b/fuzz/fuzz_targets/display_does_not_panic.rs @@ -8,14 +8,26 @@ fuzz_target!(|data: &[u8]| { let x86_16_decoder = yaxpeax_x86::real_mode::InstDecoder::default(); if let Ok(inst) = x86_64_decoder.decode_slice(data) { - inst.write_to(&mut String::new()).expect("format does not panic"); + let mut out = String::new(); + inst.write_to(&mut out).expect("format does not panic"); + let mut text_buf = yaxpeax_x86::long_mode::InstructionTextBuffer::new(); + text_buf.format_inst(&inst.display_with(yaxpeax_x86::long_mode::DisplayStyle::Intel)).expect("can format"); + assert_eq!(text_buf.text_str(), out); }; if let Ok(inst) = x86_32_decoder.decode_slice(data) { - inst.write_to(&mut String::new()).expect("format does not panic"); + let mut out = String::new(); + inst.write_to(&mut out).expect("format does not panic"); + let mut text_buf = yaxpeax_x86::protected_mode::InstructionTextBuffer::new(); + text_buf.format_inst(&inst.display_with(yaxpeax_x86::protected_mode::DisplayStyle::Intel)).expect("can format"); + assert_eq!(text_buf.text_str(), out); }; if let Ok(inst) = x86_16_decoder.decode_slice(data) { - inst.write_to(&mut String::new()).expect("format does not panic"); + let mut out = String::new(); + inst.write_to(&mut out).expect("format does not panic"); + let mut text_buf = yaxpeax_x86::real_mode::InstructionTextBuffer::new(); + text_buf.format_inst(&inst.display_with(yaxpeax_x86::real_mode::DisplayStyle::Intel)).expect("can format"); + assert_eq!(text_buf.text_str(), out); }; }); diff --git a/src/protected_mode/display.rs b/src/protected_mode/display.rs index 1db34cb..0291246 100644 --- a/src/protected_mode/display.rs +++ b/src/protected_mode/display.rs @@ -2942,3 +2942,107 @@ impl<'a, F: DisplaySink> super::OperandVisitor for RelativeBranchPrinter<'a, F> Ok(false) } } + +#[cfg(feature="alloc")] +mod buffer_sink { + use core::fmt; + use super::super::{DisplayStyle, InstructionDisplayer}; + use super::{contextualize_c, contextualize_intel}; + + /// helper to format `amd64` instructions with highest throughput and least configuration. this is + /// functionally a buffer for one x86 instruction's text. + /// + /// ### when to use this over `fmt::Display`? + /// + /// `fmt::Display` is a fair choice in most cases. in some cases, `InstructionTextBuffer` may + /// support formatting options that may be difficult to configure for a `Display` impl. + /// additionally, `InstructionTextBuffer` may be able to specialize more effectively where + /// `fmt::Display`, writing to a generic `fmt::Write`, may not. + /// + /// if your use case for `yaxpeax-x86` involves being bounded on the speed of disassembling and + /// formatting instructions, [`InstructionTextBuffer::format_inst`] has been measured as up to 11% + /// faster than an equivalent `write!(buf, "{}", inst)`. + /// + /// `InstructionTextBuffer` involves internal allocations; if your use case for `yaxpeax-x86` + /// requires allocations never occurring, it is not an appropriate tool. + /// + /// ### example + /// + /// ``` + /// use yaxpeax_x86::long_mode::InstDecoder; + /// use yaxpeax_x86::long_mode::InstructionTextBuffer; + /// use yaxpeax_x86::long_mode::DisplayStyle; + /// + /// let bytes = &[0x33, 0xc0]; + /// let inst = InstDecoder::default().decode_slice(bytes).expect("can decode"); + /// let mut text_buf = InstructionTextBuffer::new(); + /// assert_eq!( + /// text_buf.format_inst(&inst.display_with(DisplayStyle::Intel)).expect("can format"), + /// "xor eax, eax" + /// ); + /// + /// // or, getting the formatted instruction with `text_str`: + /// assert_eq!( + /// text_buf.text_str(), + /// "xor eax, eax" + /// ); + /// ``` + pub struct InstructionTextBuffer { + content: alloc::string::String, + } + + impl InstructionTextBuffer { + /// create an `InstructionTextBuffer` with default settings. `InstructionTextBuffer`'s default + /// settings format instructions identically to their corresponding `fmt::Display`. + pub fn new() -> Self { + let mut buf = alloc::string::String::new(); + // TODO: move 512 out to a MAX_INSTRUCTION_LEN const and appropriate justification (and + // fuzzing and ..) + buf.reserve(512); + Self { + content: buf, + } + } + + /// format `inst` into this buffer. returns a borrow of that same internal buffer for convenience. + /// + /// this clears and reuses an internal buffer; if an instruction had been previously formatted + /// through this buffer, it will be overwritten. + pub fn format_inst<'buf, 'instr>(&'buf mut self, display: &InstructionDisplayer<'instr>) -> Result<&'buf str, fmt::Error> { + // Safety: this sink is used to format exactly one instruction and then dropped. it can + // never escape `format_inst`. + let mut handle = unsafe { self.write_handle() }; + + match display.style { + DisplayStyle::Intel => { + contextualize_intel(&display.instr, &mut handle)?; + } + DisplayStyle::C => { + contextualize_c(&display.instr, &mut handle)?; + } + } + + Ok(self.text_str()) + } + + /// return a borrow of the internal buffer. if an instruction has been formatted, the + /// returned `&str` contains that instruction's buffered text. + pub fn text_str(&self) -> &str { + self.content.as_str() + } + + /// do the necessary bookkeeping and provide an `InstructionTextSink` to write an instruction + /// into. + /// + /// SAFETY: callers must print at most one instruction into this handle. + unsafe fn write_handle(&mut self) -> yaxpeax_arch::display::InstructionTextSink { + self.content.clear(); + // Safety: `content` was just cleared, so writing begins at the start of the buffer. + // `content`is large enough to hold a fully-formatted instruction (see + // `InstructionTextBuffer::new`). + yaxpeax_arch::display::InstructionTextSink::new(&mut self.content) + } + } +} +#[cfg(feature="alloc")] +pub use buffer_sink::InstructionTextBuffer; diff --git a/src/protected_mode/mod.rs b/src/protected_mode/mod.rs index 38f8310..6d9031a 100644 --- a/src/protected_mode/mod.rs +++ b/src/protected_mode/mod.rs @@ -8,6 +8,8 @@ pub use crate::MemoryAccessSize; #[cfg(feature = "fmt")] pub use self::display::{DisplayStyle, InstructionDisplayer}; +#[cfg(all(feature = "fmt", feature = "alloc"))] +pub use self::display::InstructionTextBuffer; use core::cmp::PartialEq; use crate::safer_unchecked::unreachable_kinda_unchecked as unreachable_unchecked; diff --git a/src/real_mode/display.rs b/src/real_mode/display.rs index 90b4f0a..fdd51cf 100644 --- a/src/real_mode/display.rs +++ b/src/real_mode/display.rs @@ -2942,3 +2942,107 @@ impl<'a, F: DisplaySink> super::OperandVisitor for RelativeBranchPrinter<'a, F> Ok(false) } } + +#[cfg(feature="alloc")] +mod buffer_sink { + use core::fmt; + use super::super::{DisplayStyle, InstructionDisplayer}; + use super::{contextualize_c, contextualize_intel}; + + /// helper to format `amd64` instructions with highest throughput and least configuration. this is + /// functionally a buffer for one x86 instruction's text. + /// + /// ### when to use this over `fmt::Display`? + /// + /// `fmt::Display` is a fair choice in most cases. in some cases, `InstructionTextBuffer` may + /// support formatting options that may be difficult to configure for a `Display` impl. + /// additionally, `InstructionTextBuffer` may be able to specialize more effectively where + /// `fmt::Display`, writing to a generic `fmt::Write`, may not. + /// + /// if your use case for `yaxpeax-x86` involves being bounded on the speed of disassembling and + /// formatting instructions, [`InstructionTextBuffer::format_inst`] has been measured as up to 11% + /// faster than an equivalent `write!(buf, "{}", inst)`. + /// + /// `InstructionTextBuffer` involves internal allocations; if your use case for `yaxpeax-x86` + /// requires allocations never occurring, it is not an appropriate tool. + /// + /// ### example + /// + /// ``` + /// use yaxpeax_x86::long_mode::InstDecoder; + /// use yaxpeax_x86::long_mode::InstructionTextBuffer; + /// use yaxpeax_x86::long_mode::DisplayStyle; + /// + /// let bytes = &[0x33, 0xc0]; + /// let inst = InstDecoder::default().decode_slice(bytes).expect("can decode"); + /// let mut text_buf = InstructionTextBuffer::new(); + /// assert_eq!( + /// text_buf.format_inst(&inst.display_with(DisplayStyle::Intel)).expect("can format"), + /// "xor eax, eax" + /// ); + /// + /// // or, getting the formatted instruction with `text_str`: + /// assert_eq!( + /// text_buf.text_str(), + /// "xor eax, eax" + /// ); + /// ``` + pub struct InstructionTextBuffer { + content: alloc::string::String, + } + + impl InstructionTextBuffer { + /// create an `InstructionTextBuffer` with default settings. `InstructionTextBuffer`'s default + /// settings format instructions identically to their corresponding `fmt::Display`. + pub fn new() -> Self { + let mut buf = alloc::string::String::new(); + // TODO: move 512 out to a MAX_INSTRUCTION_LEN const and appropriate justification (and + // fuzzing and ..) + buf.reserve(512); + Self { + content: buf, + } + } + + /// format `inst` into this buffer. returns a borrow of that same internal buffer for convenience. + /// + /// this clears and reuses an internal buffer; if an instruction had been previously formatted + /// through this buffer, it will be overwritten. + pub fn format_inst<'buf, 'instr>(&'buf mut self, display: &InstructionDisplayer<'instr>) -> Result<&'buf str, fmt::Error> { + // Safety: this sink is used to format exactly one instruction and then dropped. it can + // never escape `format_inst`. + let mut handle = unsafe { self.write_handle() }; + + match display.style { + DisplayStyle::Intel => { + contextualize_intel(&display.instr, &mut handle)?; + } + DisplayStyle::C => { + contextualize_c(&display.instr, &mut handle)?; + } + } + + Ok(self.text_str()) + } + + /// return a borrow of the internal buffer. if an instruction has been formatted, the + /// returned `&str` contains that instruction's buffered text. + pub fn text_str(&self) -> &str { + self.content.as_str() + } + + /// do the necessary bookkeeping and provide an `InstructionTextSink` to write an instruction + /// into. + /// + /// SAFETY: callers must print at most one instruction into this handle. + unsafe fn write_handle(&mut self) -> yaxpeax_arch::display::InstructionTextSink { + self.content.clear(); + // Safety: `content` was just cleared, so writing begins at the start of the buffer. + // `content`is large enough to hold a fully-formatted instruction (see + // `InstructionTextBuffer::new`). + yaxpeax_arch::display::InstructionTextSink::new(&mut self.content) + } + } +} +#[cfg(feature="alloc")] +pub use buffer_sink::InstructionTextBuffer; diff --git a/src/real_mode/mod.rs b/src/real_mode/mod.rs index b1aaf7d..162b380 100644 --- a/src/real_mode/mod.rs +++ b/src/real_mode/mod.rs @@ -8,6 +8,8 @@ pub use crate::MemoryAccessSize; #[cfg(feature = "fmt")] pub use self::display::{DisplayStyle, InstructionDisplayer}; +#[cfg(all(feature = "fmt", feature = "alloc"))] +pub use self::display::InstructionTextBuffer; use core::cmp::PartialEq; use crate::safer_unchecked::unreachable_kinda_unchecked as unreachable_unchecked; |