[tarantool-patches] [PATCH] Fix memory corruption while vinyl index rebuild
Vladimir Davydov
vdavydov.dev at gmail.com
Wed Jan 9 12:41:38 MSK 2019
The issue's naturally worth fixing, but I don't think that it could
actually lead to a memory corruption, because the code doesn't yield.
AFAIU you stumbled upon it just by looking at the code, not that you
encountered a crash, right?
On Sat, Jan 05, 2019 at 12:18:01AM +0300, Georgy Kirichenko wrote:
> Reclaim memory used while previous page recovery not the last one.
> There is no specific test case.
>
> Fixes: 3920
Closes #3920
Fixes commit a14d5e4e12ed2 ("vinyl: rebuild .index in force-recovery
mode").
>
> Issue: https://github.com/tarantool/tarantool/issues/3920
> Branch: https://github.com/tarantool/tarantool/tree/g.kirichenko/gh-3920-rebuild-mem-corruption
Should go after --
'Issue' and 'Branch' prefixes are useless.
> diff --git a/src/box/vy_run.c b/src/box/vy_run.c
> index c448a449d..d82f1e372 100644
> --- a/src/box/vy_run.c
> +++ b/src/box/vy_run.c
> @@ -2382,6 +2382,7 @@ vy_run_rebuild_index(struct vy_run *run, const char *dir,
>
> off_t page_offset, next_page_offset = xlog_cursor_pos(&cursor);
> while ((rc = xlog_cursor_next_tx(&cursor)) == 0) {
> + region_truncate(region, mem_used);
> page_offset = next_page_offset;
> next_page_offset = xlog_cursor_pos(&cursor);
>
> @@ -2441,7 +2442,6 @@ vy_run_rebuild_index(struct vy_run *run, const char *dir,
> info->row_index_offset = page_row_index_offset;
> ++run->info.page_count;
> vy_run_acct_page(run, info);
> - region_truncate(region, mem_used);
> }
>
> if (prev_tuple != NULL) {
Pushed to 2.1 and 1.10.
More information about the Tarantool-patches
mailing list