diff options
author | Linus Torvalds <torvalds@linux-foundation.org> | 2021-05-03 11:19:54 -0700 |
---|---|---|
committer | Linus Torvalds <torvalds@linux-foundation.org> | 2021-05-03 11:19:54 -0700 |
commit | 9b1f61d5d73d550a20dd79b9a17b6bb05a8f9307 (patch) | |
tree | b1b3c9677f586971212fee1c38500ff5d2b6773a /kernel/trace/trace_events.c | |
parent | 6f8ee8d36d076b517028b60911877e27bc1d8363 (diff) | |
parent | aafe104aa9096827a429bc1358f8260ee565b7cc (diff) | |
download | linux-9b1f61d5d73d550a20dd79b9a17b6bb05a8f9307.tar.gz |
Merge tag 'trace-v5.13' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace
Pull tracing updates from Steven Rostedt: "New feature: - A new "func-no-repeats" option in tracefs/options directory. When set the function tracer will detect if the current function being traced is the same as the previous one, and instead of recording it, it will keep track of the number of times that the function is repeated in a row. And when another function is recorded, it will write a new event that shows the function that repeated, the number of times it repeated and the time stamp of when the last repeated function occurred. Enhancements: - In order to implement the above "func-no-repeats" option, the ring buffer timestamp can now give the accurate timestamp of the event as it is being recorded, instead of having to record an absolute timestamp for all events. This helps the histogram code which no longer needs to waste ring buffer space. - New validation logic to make sure all trace events that access dereferenced pointers do so in a safe way, and will warn otherwise. Fixes: - No longer limit the PIDs of tasks that are recorded for "saved_cmdlines" to PID_MAX_DEFAULT (32768), as systemd now allows for a much larger range. This caused the mapping of PIDs to the task names to be dropped for all tasks with a PID greater than 32768. - Change trace_clock_global() to never block. This caused a deadlock. Clean ups: - Typos, prototype fixes, and removing of duplicate or unused code. - Better management of ftrace_page allocations" * tag 'trace-v5.13' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace: (32 commits) tracing: Restructure trace_clock_global() to never block tracing: Map all PIDs to command lines ftrace: Reuse the output of the function tracer for func_repeats tracing: Add "func_no_repeats" option for function tracing tracing: Unify the logic for function tracing options tracing: Add method for recording "func_repeats" events tracing: Add "last_func_repeats" to struct trace_array tracing: Define new ftrace event "func_repeats" tracing: Define static void trace_print_time() ftrace: Simplify the calculation of page number for ftrace_page->records some more ftrace: Store the order of pages allocated in ftrace_page tracing: Remove unused argument from "ring_buffer_time_stamp() tracing: Remove duplicate struct declaration in trace_events.h tracing: Update create_system_filter() kernel-doc comment tracing: A minor cleanup for create_system_filter() kernel: trace: Mundane typo fixes in the file trace_events_filter.c tracing: Fix various typos in comments scripts/recordmcount.pl: Make vim and emacs indent the same scripts/recordmcount.pl: Make indent spacing consistent tracing: Add a verifier to check string pointers for trace events ...
Diffstat (limited to 'kernel/trace/trace_events.c')
-rw-r--r-- | kernel/trace/trace_events.c | 214 |
1 files changed, 212 insertions, 2 deletions
diff --git a/kernel/trace/trace_events.c b/kernel/trace/trace_events.c index a3563afd412d..80e96989770e 100644 --- a/kernel/trace/trace_events.c +++ b/kernel/trace/trace_events.c @@ -217,6 +217,214 @@ int trace_event_get_offsets(struct trace_event_call *call) return tail->offset + tail->size; } +/* + * Check if the referenced field is an array and return true, + * as arrays are OK to dereference. + */ +static bool test_field(const char *fmt, struct trace_event_call *call) +{ + struct trace_event_fields *field = call->class->fields_array; + const char *array_descriptor; + const char *p = fmt; + int len; + + if (!(len = str_has_prefix(fmt, "REC->"))) + return false; + fmt += len; + for (p = fmt; *p; p++) { + if (!isalnum(*p) && *p != '_') + break; + } + len = p - fmt; + + for (; field->type; field++) { + if (strncmp(field->name, fmt, len) || + field->name[len]) + continue; + array_descriptor = strchr(field->type, '['); + /* This is an array and is OK to dereference. */ + return array_descriptor != NULL; + } + return false; +} + +/* + * Examine the print fmt of the event looking for unsafe dereference + * pointers using %p* that could be recorded in the trace event and + * much later referenced after the pointer was freed. Dereferencing + * pointers are OK, if it is dereferenced into the event itself. + */ +static void test_event_printk(struct trace_event_call *call) +{ + u64 dereference_flags = 0; + bool first = true; + const char *fmt, *c, *r, *a; + int parens = 0; + char in_quote = 0; + int start_arg = 0; + int arg = 0; + int i; + + fmt = call->print_fmt; + + if (!fmt) + return; + + for (i = 0; fmt[i]; i++) { + switch (fmt[i]) { + case '\\': + i++; + if (!fmt[i]) + return; + continue; + case '"': + case '\'': + /* + * The print fmt starts with a string that + * is processed first to find %p* usage, + * then after the first string, the print fmt + * contains arguments that are used to check + * if the dereferenced %p* usage is safe. + */ + if (first) { + if (fmt[i] == '\'') + continue; + if (in_quote) { + arg = 0; + first = false; + /* + * If there was no %p* uses + * the fmt is OK. + */ + if (!dereference_flags) + return; + } + } + if (in_quote) { + if (in_quote == fmt[i]) + in_quote = 0; + } else { + in_quote = fmt[i]; + } + continue; + case '%': + if (!first || !in_quote) + continue; + i++; + if (!fmt[i]) + return; + switch (fmt[i]) { + case '%': + continue; + case 'p': + /* Find dereferencing fields */ + switch (fmt[i + 1]) { + case 'B': case 'R': case 'r': + case 'b': case 'M': case 'm': + case 'I': case 'i': case 'E': + case 'U': case 'V': case 'N': + case 'a': case 'd': case 'D': + case 'g': case 't': case 'C': + case 'O': case 'f': + if (WARN_ONCE(arg == 63, + "Too many args for event: %s", + trace_event_name(call))) + return; + dereference_flags |= 1ULL << arg; + } + break; + default: + { + bool star = false; + int j; + + /* Increment arg if %*s exists. */ + for (j = 0; fmt[i + j]; j++) { + if (isdigit(fmt[i + j]) || + fmt[i + j] == '.') + continue; + if (fmt[i + j] == '*') { + star = true; + continue; + } + if ((fmt[i + j] == 's') && star) + arg++; + break; + } + break; + } /* default */ + + } /* switch */ + arg++; + continue; + case '(': + if (in_quote) + continue; + parens++; + continue; + case ')': + if (in_quote) + continue; + parens--; + if (WARN_ONCE(parens < 0, + "Paren mismatch for event: %s\narg='%s'\n%*s", + trace_event_name(call), + fmt + start_arg, + (i - start_arg) + 5, "^")) + return; + continue; + case ',': + if (in_quote || parens) + continue; + i++; + while (isspace(fmt[i])) + i++; + start_arg = i; + if (!(dereference_flags & (1ULL << arg))) + goto next_arg; + + /* Find the REC-> in the argument */ + c = strchr(fmt + i, ','); + r = strstr(fmt + i, "REC->"); + if (r && (!c || r < c)) { + /* + * Addresses of events on the buffer, + * or an array on the buffer is + * OK to dereference. + * There's ways to fool this, but + * this is to catch common mistakes, + * not malicious code. + */ + a = strchr(fmt + i, '&'); + if ((a && (a < r)) || test_field(r, call)) + dereference_flags &= ~(1ULL << arg); + } + next_arg: + i--; + arg++; + } + } + + /* + * If you triggered the below warning, the trace event reported + * uses an unsafe dereference pointer %p*. As the data stored + * at the trace event time may no longer exist when the trace + * event is printed, dereferencing to the original source is + * unsafe. The source of the dereference must be copied into the + * event itself, and the dereference must access the copy instead. + */ + if (WARN_ON_ONCE(dereference_flags)) { + arg = 1; + while (!(dereference_flags & 1)) { + dereference_flags >>= 1; + arg++; + } + pr_warn("event %s has unsafe dereference of argument %d\n", + trace_event_name(call), arg); + pr_warn("print_fmt: %s\n", fmt); + } +} + int trace_event_raw_init(struct trace_event_call *call) { int id; @@ -225,6 +433,8 @@ int trace_event_raw_init(struct trace_event_call *call) if (!id) return -ENODEV; + test_event_printk(call); + return 0; } EXPORT_SYMBOL_GPL(trace_event_raw_init); @@ -2436,7 +2646,7 @@ void trace_event_eval_update(struct trace_eval_map **map, int len) } /* - * Since calls are grouped by systems, the likelyhood that the + * Since calls are grouped by systems, the likelihood that the * next call in the iteration belongs to the same system as the * previous call is high. As an optimization, we skip searching * for a map[] that matches the call's system if the last call @@ -2496,7 +2706,7 @@ __trace_add_new_event(struct trace_event_call *call, struct trace_array *tr) } /* - * Just create a decriptor for early init. A descriptor is required + * Just create a descriptor for early init. A descriptor is required * for enabling events at boot. We want to enable events before * the filesystem is initialized. */ |