On Mon, 15 Jul 2024 16:51:41 +0200 Petr Pavlu <[email protected]> wrote:
I finally got around to looking at this. Sorry for the delay. > The function rb_check_pages() validates the integrity of a specified > per-CPU tracing ring buffer. It does so by traversing the underlying > linked list and checking its next and prev links. > > To guarantee that the list isn't modified during the check, a caller > typically needs to take cpu_buffer->reader_lock. This prevents the check > from running concurrently, for example, with a potential reader which > can make the list temporarily inconsistent when swapping its old reader > page into the buffer. > > A problem with this approach is that the time when interrupts are > disabled is non-deterministic, dependent on the ring buffer size. This > particularly affects PREEMPT_RT because the reader_lock is a raw > spinlock which doesn't become sleepable on PREEMPT_RT kernels. > > Modify the check so it still attempts to traverse the entire list, but > gives up the reader_lock between checking individual pages. Introduce > for this purpose a new variable ring_buffer_per_cpu.cnt which is bumped > any time the list is modified. The value is used by rb_check_pages() to > detect such a change and restart the check. > > Signed-off-by: Petr Pavlu <[email protected]> > --- > > This is a follow-up to the discussion about improving this check [1]. > > Changes since v2 [2]: > * Call the new generation variable as ring_buffer_per_cpu.cnt. > > Changes since v1 [3]: > * Correct the case when rb_check_pages() is invoked concurrently with > the resize code which modifies the list. Introduce > ring_buffer_per_cpu.pages_era for this purpose. > > [1] > https://lore.kernel.org/linux-trace-kernel/[email protected]/ > [2] > https://lore.kernel.org/linux-trace-kernel/[email protected]/ > [2] > https://lore.kernel.org/linux-trace-kernel/[email protected]/ > > kernel/trace/ring_buffer.c | 98 ++++++++++++++++++++++++++++---------- > 1 file changed, 72 insertions(+), 26 deletions(-) > > diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c > index 28853966aa9a..0279b4ee0b03 100644 > --- a/kernel/trace/ring_buffer.c > +++ b/kernel/trace/ring_buffer.c > @@ -462,6 +462,8 @@ struct ring_buffer_per_cpu { > unsigned long nr_pages; > unsigned int current_context; > struct list_head *pages; > + /* pages generation counter, incremented when the list changes */ > + unsigned long cnt; > struct buffer_page *head_page; /* read from head */ > struct buffer_page *tail_page; /* write to tail */ > struct buffer_page *commit_page; /* committed pages */ > @@ -1454,40 +1456,87 @@ static void rb_check_bpage(struct ring_buffer_per_cpu > *cpu_buffer, > RB_WARN_ON(cpu_buffer, val & RB_FLAG_MASK); > } > > +static bool rb_check_links(struct ring_buffer_per_cpu *cpu_buffer, > + struct list_head *list) > +{ > + if (RB_WARN_ON(cpu_buffer, > + rb_list_head(rb_list_head(list->next)->prev) != list)) > + return false; > + > + if (RB_WARN_ON(cpu_buffer, > + rb_list_head(rb_list_head(list->prev)->next) != list)) > + return false; > + > + return true; > +} > + > /** > * rb_check_pages - integrity check of buffer pages > * @cpu_buffer: CPU buffer with pages to test > * > * As a safety measure we check to make sure the data pages have not > * been corrupted. > - * > - * Callers of this function need to guarantee that the list of pages doesn't > get > - * modified during the check. In particular, if it's possible that the > function > - * is invoked with concurrent readers which can swap in a new reader page > then > - * the caller should take cpu_buffer->reader_lock. > */ > static void rb_check_pages(struct ring_buffer_per_cpu *cpu_buffer) > { > - struct list_head *head = rb_list_head(cpu_buffer->pages); > - struct list_head *tmp; > + struct list_head *head, *tmp; > + unsigned long buffer_cnt; > + unsigned long flags; > + int nr_loops = 0; > > - if (RB_WARN_ON(cpu_buffer, > - rb_list_head(rb_list_head(head->next)->prev) != head)) > + /* > + * Walk the linked list underpinning the ring buffer and validate all > + * its next and prev links. > + * > + * The check acquires the reader_lock to avoid concurrent processing > + * with code that could be modifying the list. However, the lock cannot > + * be held for the entire duration of the walk, as this would make the > + * time when interrupts are disabled non-deterministic, dependent on the > + * ring buffer size. Therefore, the code releases and re-acquires the > + * lock after checking each page. The ring_buffer_per_cpu.cnt variable > + * is then used to detect if the list was modified while the lock was > + * not held, in which case the check needs to be restarted. > + * > + * The code attempts to perform the check at most three times before > + * giving up. This is acceptable because this is only a self-validation > + * to detect problems early on. In practice, the list modification > + * operations are fairly spaced, and so this check typically succeeds at > + * most on the second try. > + */ > +again: > + if (++nr_loops > 3) > return; > > - if (RB_WARN_ON(cpu_buffer, > - rb_list_head(rb_list_head(head->prev)->next) != head)) > - return; > + raw_spin_lock_irqsave(&cpu_buffer->reader_lock, flags); > + head = rb_list_head(cpu_buffer->pages); > + if (!rb_check_links(cpu_buffer, head)) > + goto out_locked; > + buffer_cnt = cpu_buffer->cnt; > + tmp = head; > + raw_spin_unlock_irqrestore(&cpu_buffer->reader_lock, flags); > > - for (tmp = rb_list_head(head->next); tmp != head; tmp = > rb_list_head(tmp->next)) { > - if (RB_WARN_ON(cpu_buffer, > - rb_list_head(rb_list_head(tmp->next)->prev) != > tmp)) > - return; > + while (true) { > + raw_spin_lock_irqsave(&cpu_buffer->reader_lock, flags); > > - if (RB_WARN_ON(cpu_buffer, > - rb_list_head(rb_list_head(tmp->prev)->next) != > tmp)) > - return; > + if (buffer_cnt != cpu_buffer->cnt) { > + /* The list was updated, try again. */ > + raw_spin_unlock_irqrestore(&cpu_buffer->reader_lock, > flags); > + goto again; > + } > + > + tmp = rb_list_head(tmp->next); > + if (tmp == head) > + /* The iteration circled back, all is done. */ > + goto out_locked; > + > + if (!rb_check_links(cpu_buffer, tmp)) > + goto out_locked; > + > + raw_spin_unlock_irqrestore(&cpu_buffer->reader_lock, flags); > } > + > +out_locked: > + raw_spin_unlock_irqrestore(&cpu_buffer->reader_lock, flags); > } > > static int __rb_allocate_pages(struct ring_buffer_per_cpu *cpu_buffer, > @@ -1874,6 +1923,7 @@ rb_remove_pages(struct ring_buffer_per_cpu *cpu_buffer, > unsigned long nr_pages) > > /* make sure pages points to a valid page in the ring buffer */ > cpu_buffer->pages = next_page; > + cpu_buffer->cnt++; > > /* update head page */ > if (head_bit) > @@ -1980,6 +2030,7 @@ rb_insert_pages(struct ring_buffer_per_cpu *cpu_buffer) > * pointer to point to end of list > */ > head_page->prev = last_page; > + cpu_buffer->cnt++; > success = true; > break; > } > @@ -2215,12 +2266,8 @@ int ring_buffer_resize(struct trace_buffer *buffer, > unsigned long size, > */ > synchronize_rcu(); > for_each_buffer_cpu(buffer, cpu) { > - unsigned long flags; > - > cpu_buffer = buffer->buffers[cpu]; > - raw_spin_lock_irqsave(&cpu_buffer->reader_lock, flags); > rb_check_pages(cpu_buffer); > - raw_spin_unlock_irqrestore(&cpu_buffer->reader_lock, > flags); > } > atomic_dec(&buffer->record_disabled); > } > @@ -4611,6 +4658,7 @@ rb_get_reader_page(struct ring_buffer_per_cpu > *cpu_buffer) > rb_list_head(reader->list.next)->prev = &cpu_buffer->reader_page->list; > rb_inc_page(&cpu_buffer->head_page); > > + cpu_buffer->cnt++; > local_inc(&cpu_buffer->pages_read); > > /* Finally update the reader page to the new head */ > @@ -5150,12 +5198,9 @@ void > ring_buffer_read_finish(struct ring_buffer_iter *iter) > { > struct ring_buffer_per_cpu *cpu_buffer = iter->cpu_buffer; > - unsigned long flags; > > /* Use this opportunity to check the integrity of the ring buffer. */ > - raw_spin_lock_irqsave(&cpu_buffer->reader_lock, flags); > rb_check_pages(cpu_buffer); > - raw_spin_unlock_irqrestore(&cpu_buffer->reader_lock, flags); > > atomic_dec(&cpu_buffer->resize_disabled); > kfree(iter->event); > @@ -6050,6 +6095,7 @@ int ring_buffer_subbuf_order_set(struct trace_buffer > *buffer, int order) > cpu_buffer->pages = cpu_buffer->new_pages.next; > cpu_buffer->new_pages.next->prev = cpu_buffer->new_pages.prev; > cpu_buffer->new_pages.prev->next = cpu_buffer->new_pages.next; > + cpu_buffer->cnt++; Everything looks good but this cnt update isn't protected. Probably doesn't matter but it feels like it should be. Perhaps we can just convert it to an atomic? -- Steve > > /* Clear the new_pages list */ > INIT_LIST_HEAD(&cpu_buffer->new_pages); > > base-commit: 0c3836482481200ead7b416ca80c68a29cfdaabd
