Summary

Fixes the PHP memory exhaustion issue described in pfrest#806.

Log endpoints (/api/v2/status/logs/*) load all log files - including every rotated/compressed copy - into memory before Model::read_all() applies pagination via array_slice(). Even ?limit=1 triggers loading the entire log history, causing OOM at 512MB on systems with moderate log volumes.

This PR passes the pagination limit from Model::read_all() down through get_internal_objects() to the log-reading callables, so they can stop reading early once enough lines are collected.

Changes

LogFileModelTraits.inc - the core fix:

• read_log() gains an optional $limit parameter
• When $limit > 0: reads newest files first, stops when enough lines are collected
• Plain text files: reverse-reads from EOF via fseek() - O(limit) memory
• Gzip files: streams line-by-line with a ring buffer - O(limit) memory
• Bzip2/xz: full decompression (no streaming API), but truncated immediately and skipped entirely if newer files already satisfy the limit
• gather_log_filepaths() now returns newest-first ordering with explicit sort
• When $limit = 0 (default): preserves existing behavior exactly

Model.inc - plumbing:

• get_internal_objects() gains optional int $limit parameter
• Uses ReflectionMethod to check if the callable accepts a $limit param before passing it (backward-compatible with callables that don't)
• read_all() passes $limit + $offset as the max needed hint

All 6 log models - trivial forwarding:

• FirewallLog, SystemLog, DHCPLog, AuthLog, OpenVPNLog, RESTAPILog
• Each callable gains int $limit = 0 and forwards it to read_log()

Performance impact

For ?limit=50 - the common case - this typically reads only the tail of the current log file and never touches the compressed rotated copies, reducing memory from O(total_log_size) to O(limit).

Backward compatibility

• API response format unchanged (same {'text': ...} entries)
• Unbounded reads follow the exact same code path
• Callables without a $limit parameter continue to work (reflection check)
• query() with filter parameters still does full reads (required for correctness)

Test plan

• All 5 existing LogFileModelTraits test cases pass unchanged (unbounded reads)
• 5 new test cases for bounded reads:
  • Bounded read returns correct newest entries across multiple files
  • Bounded read from current file only when limit < current file lines
  • Bounded read with gzip compressed rotated logs
  • Unbounded read (limit=0) returns all entries (backward compat)
  • Bounded read with limit > total lines returns everything without error
• Manual test: GET /api/v2/status/logs/firewall?limit=50 no longer OOMs on systems with large log files
• Manual test: GET /api/v2/status/logs/firewall (no limit) still returns all entries

Ref: pfrest#806

🤖 Generated with Claude Code