Skip to content

Commit

Permalink
Adds diff to accounts-hash-cache-tool (anza-xyz#1772)
Browse files Browse the repository at this point in the history
  • Loading branch information
@brooksprumo @samkim-crypto
brooksprumo authored and samkim-crypto committed Jul 31, 2024
1 parent 559ccb2 commit c6d5e50
Show file tree
Hide file tree
Showing 2 changed files with 241 additions and 53 deletions.
2 changes: 1 addition & 1 deletion accounts-db/accounts-hash-cache-tool/Cargo.toml
View file
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
[package]
name = "agave-accounts-hash-cache-tool"
description = "Tool to inspect accounts hash cache files"
description = "Tool for accounts hash cache files"
publish = false
version = { workspace = true }
authors = { workspace = true }
Expand Down
292 changes: 240 additions & 52 deletions accounts-db/accounts-hash-cache-tool/src/main.rs
View file
Original file line number Diff line number Diff line change
@@ -1,75 +1,91 @@
use {
bytemuck::Zeroable as _,
clap::{crate_description, crate_name, value_t_or_exit, App, Arg},
clap::{
crate_description, crate_name, value_t_or_exit, App, AppSettings, Arg, ArgMatches,
SubCommand,
},
solana_accounts_db::{CacheHashDataFileEntry, CacheHashDataFileHeader},
std::{
collections::HashMap,
fs::File,
io::{self, BufReader, Read as _},
mem::size_of,
num::Saturating,
path::Path,
},
};

fn main() {
let matches = App::new(crate_name!())
.about(crate_description!())
.version(solana_version::version!())
.arg(
Arg::with_name("path")
.index(1)
.takes_value(true)
.value_name("PATH")
.help("Accounts hash cache file to inspect"),
.global_setting(AppSettings::ArgRequiredElseHelp)
.global_setting(AppSettings::ColoredHelp)
.global_setting(AppSettings::InferSubcommands)
.global_setting(AppSettings::UnifiedHelpMessage)
.global_setting(AppSettings::VersionlessSubcommands)
.subcommand(
SubCommand::with_name("inspect")
.about(
"Inspect an accounts hash cache file and display \
each account's address, hash, and balance",
)
.arg(
Arg::with_name("force")
.long("force")
.takes_value(false)
.help("Continue even if sanity checks fail"),
)
.arg(
Arg::with_name("path")
.index(1)
.takes_value(true)
.value_name("PATH")
.help("Accounts hash cache file to inspect"),
),
)
.arg(
Arg::with_name("force")
.long("force")
.takes_value(false)
.help("Continue even if sanity checks fail"),
.subcommand(
SubCommand::with_name("diff")
.about("Diff two accounts hash cache files")
.arg(
Arg::with_name("path1")
.index(1)
.takes_value(true)
.value_name("PATH1")
.help("Accounts hash cache file 1 to diff"),
)
.arg(
Arg::with_name("path2")
.index(2)
.takes_value(true)
.value_name("PATH2")
.help("Accounts hash cache file 2 to diff"),
),
)
.get_matches();

let force = matches.is_present("force");
let path = value_t_or_exit!(matches, "path", String);

let file = File::open(&path).unwrap_or_else(|err| {
eprintln!("Failed to open accounts hash cache file '{path}': {err}");
std::process::exit(1);
});
let actual_file_size = file
.metadata()
.unwrap_or_else(|err| {
eprintln!("Failed to query file metadata: {err}");
std::process::exit(1);
})
.len();
let mut reader = BufReader::new(file);

let header = {
let mut header = CacheHashDataFileHeader::zeroed();
reader
.read_exact(bytemuck::bytes_of_mut(&mut header))
.unwrap_or_else(|err| {
eprintln!("Failed to read header: {err}");
std::process::exit(1);
});
header
};

// Sanity checks -- ensure the actual file size matches the expected file size
let expected_file_size = size_of::<CacheHashDataFileHeader>()
.saturating_add(size_of::<CacheHashDataFileEntry>().saturating_mul(header.count));
if actual_file_size != expected_file_size as u64 {
eprintln!(
"Failed sanitization: actual file size does not match expected file size! \
actual: {actual_file_size}, expected: {expected_file_size}",
);
if !force {
std::process::exit(1);
match matches.subcommand() {
("inspect", Some(subcommand_matches)) => do_inspect(&matches, subcommand_matches)
.map_err(|err| format!("inspection failed: {err}")),
("diff", Some(subcommand_matches)) => {
do_diff(&matches, subcommand_matches).map_err(|err| format!("diff failed: {err}"))
}
eprintln!("Forced. Continuing... Results may be incorrect.");
_ => unreachable!(),
}
.unwrap_or_else(|err| {
eprintln!("Error: {err}");
std::process::exit(1);
});
}

fn do_inspect(
_app_matches: &ArgMatches<'_>,
subcommand_matches: &ArgMatches<'_>,
) -> Result<(), String> {
let force = subcommand_matches.is_present("force");
let path = value_t_or_exit!(subcommand_matches, "path", String);
let (mut reader, header) = open_file(&path, force)
.map_err(|err| format!("failed to open accounts hash cache file '{path}': {err}"))?;
let count_width = (header.count as f64).log10().ceil() as usize;
let mut count = Saturating(0usize);
loop {
Expand All @@ -80,10 +96,13 @@ fn main() {
Err(err) => {
if err.kind() == io::ErrorKind::UnexpectedEof && count.0 == header.count {
// we've hit the expected end of the file
break;
} else {
eprintln!("Failed to read entry {count}: {err}");
return Err(format!(
"failed to read entry {count}, expected {}: {err}",
header.count,
));
}
break;
}
};
println!(
Expand All @@ -96,4 +115,173 @@ fn main() {
}

println!("actual entries: {count}, expected: {}", header.count);
Ok(())
}

fn do_diff(
_app_matches: &ArgMatches<'_>,
subcommand_matches: &ArgMatches<'_>,
) -> Result<(), String> {
let force = false; // skipping sanity checks is not supported when diffing
let path1 = value_t_or_exit!(subcommand_matches, "path1", String);
let path2 = value_t_or_exit!(subcommand_matches, "path2", String);
let (mut reader1, header1) = open_file(&path1, force)
.map_err(|err| format!("failed to open accounts hash cache file 1 '{path1}': {err}"))?;
let (mut reader2, header2) = open_file(&path2, force)
.map_err(|err| format!("failed to open accounts hash cache file 2 '{path2}': {err}"))?;
// Note: Purposely open both files before reading either one. This way, if there's an error
// opening file 2, we can bail early without having to wait for file 1 to be read completely.

// extract the entries from both files
let do_extract = |num, reader: &mut BufReader<_>, header: &CacheHashDataFileHeader| {
let mut entries = HashMap::<_, _>::default();
loop {
let mut entry = CacheHashDataFileEntry::zeroed();
let result = reader.read_exact(bytemuck::bytes_of_mut(&mut entry));
match result {
Ok(()) => {}
Err(err) => {
if err.kind() == io::ErrorKind::UnexpectedEof && entries.len() == header.count {
// we've hit the expected end of the file
break;
} else {
return Err(format!(
"failed to read entry {}, expected {}: {err}",
entries.len(),
header.count,
));
}
}
};
let CacheHashDataFileEntry {
hash,
lamports,
pubkey,
} = entry;
let old_value = entries.insert(pubkey, (hash, lamports));
if let Some(old_value) = old_value {
let new_value = entries.get(&pubkey);
return Err(format!("found duplicate pubkey in file {num}: {pubkey}, old value: {old_value:?}, new value: {new_value:?}"));
}
}
Ok(entries)
};
let entries1 = do_extract(1, &mut reader1, &header1)?;
let entries2 = do_extract(2, &mut reader2, &header2)?;

// compute the differences between the files
let do_compute = |lhs: &HashMap<_, (_, _)>, rhs: &HashMap<_, (_, _)>| {
let mut unique_entries = Vec::new();
let mut mismatch_entries = Vec::new();
for (lhs_key, lhs_value) in lhs.iter() {
if let Some(rhs_value) = rhs.get(lhs_key) {
if lhs_value != rhs_value {
mismatch_entries.push((
CacheHashDataFileEntry {
hash: lhs_value.0,
lamports: lhs_value.1,
pubkey: *lhs_key,
},
CacheHashDataFileEntry {
hash: rhs_value.0,
lamports: rhs_value.1,
pubkey: *lhs_key,
},
));
}
} else {
unique_entries.push(CacheHashDataFileEntry {
hash: lhs_value.0,
lamports: lhs_value.1,
pubkey: *lhs_key,
});
}
}
unique_entries.sort_unstable_by(|a, b| a.pubkey.cmp(&b.pubkey));
mismatch_entries.sort_unstable_by(|a, b| a.0.pubkey.cmp(&b.0.pubkey));
(unique_entries, mismatch_entries)
};
let (unique_entries1, mismatch_entries) = do_compute(&entries1, &entries2);
let (unique_entries2, _) = do_compute(&entries2, &entries1);

// display the unique entries in each file
let do_print = |entries: &[CacheHashDataFileEntry]| {
let count_width = (entries.len() as f64).log10().ceil() as usize;
if entries.is_empty() {
println!("(none)");
} else {
for (i, entry) in entries.iter().enumerate() {
println!(
"{i:count_width$}: pubkey: {:44}, hash: {:44}, lamports: {}",
entry.pubkey.to_string(),
entry.hash.0.to_string(),
entry.lamports,
);
}
}
};
println!("Unique entries in file 1:");
do_print(&unique_entries1);
println!("Unique entries in file 2:");
do_print(&unique_entries2);

println!("Mismatch values:");
let count_width = (mismatch_entries.len() as f64).log10().ceil() as usize;
if mismatch_entries.is_empty() {
println!("(none)");
} else {
for (i, (lhs, rhs)) in mismatch_entries.iter().enumerate() {
println!(
"{i:count_width$}: pubkey: {:44}, hash: {:44}, lamports: {}",
lhs.pubkey.to_string(),
lhs.hash.0.to_string(),
lhs.lamports,
);
println!(
"{i:count_width$}: file 2: {:44}, hash: {:44}, lamports: {}",
"(same)".to_string(),
rhs.hash.0.to_string(),
rhs.lamports,
);
}
}

Ok(())
}

fn open_file(
path: impl AsRef<Path>,
force: bool,
) -> Result<(BufReader<File>, CacheHashDataFileHeader), String> {
let file = File::open(path).map_err(|err| format!("{err}"))?;
let actual_file_size = file
.metadata()
.map_err(|err| format!("failed to query file metadata: {err}"))?
.len();
let mut reader = BufReader::new(file);

let header = {
let mut header = CacheHashDataFileHeader::zeroed();
reader
.read_exact(bytemuck::bytes_of_mut(&mut header))
.map_err(|err| format!("failed to read header: {err}"))?;
header
};

// Sanity checks -- ensure the actual file size matches the expected file size
let expected_file_size = size_of::<CacheHashDataFileHeader>()
.saturating_add(size_of::<CacheHashDataFileEntry>().saturating_mul(header.count));
if actual_file_size != expected_file_size as u64 {
let err_msg = format!(
"failed sanitization: actual file size does not match expected file size! \
actual: {actual_file_size}, expected: {expected_file_size}",
);
if force {
eprintln!("Warning: {err_msg}\nForced. Continuing... Results may be incorrect.");
} else {
return Err(err_msg);
}
}

Ok((reader, header))
}

0 comments on commit c6d5e50

Please sign in to comment.