1berry/MariaDB-server
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Fixup the parent commit for MSAN and Valgrind

Browse Source 
commit 484931325e7bcecddc6daf1a55c008ddd67497e4 was a necessary
fix for the buffer pool resizing tests in 10.5 in
AddressSanitizer. However, that change would break the tests
innodb.innodb_buffer_pool_resize and
innodb.innodb_buffer_pool_resize_with_chunks
when run in MemorySanitizer, or presumably in Valgrind as well.
(Those tests run "forever" in Valgrind.)

buf_pool_resize(): Cancel the effect of MEM_NOACCESS() in Valgrind
and ASAN. In MSAN, MEM_NOACCESS() is a no-op, and hence we must do
nothing special here.

MEM_MAKE_ADDRESSABLE() would declare the memory contents undefined.
In this particular case, we must actually declare the contents
defined for Valgrind.
This commit is contained in:
Marko Mäkelä 2020-07-04 22:02:30 +03:00
parent 484931325e
commit 453dc4b300
1 changed files with 16 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

19
storage/innobase/buf/buf0buf.cc
View File

@ -1622,7 +1622,7 @@ buf_chunk_init(
return(NULL);
}
MEM_MAKE_ADDRESSABLE(chunk->mem, chunk->mem_size());
MEM_MAKE_ADDRESSABLE(chunk->mem, chunk->mem_size());
#ifdef HAVE_LIBNUMA
if (srv_numa_interleave) {
@ -2910,8 +2910,21 @@ withdraw_retry:
while (chunk < echunk) {
buf_block_t* block = chunk->blocks;
MEM_MAKE_ADDRESSABLE(chunk->mem,
chunk->mem_size());
/* buf_LRU_block_free_non_file_page()
invokes MEM_NOACCESS() on any blocks
that are in free_list. We must
cancel the effect of that. In MemorySanitizer,
MEM_NOACCESS() is no-op, so we must not do
anything special for it here. */
#ifdef HAVE_valgrind
# if !__has_feature(memory_sanitizer)
MEM_MAKE_DEFINED(chunk->mem,
chunk->mem_size());
# endif
#else
MEM_MAKE_ADDRESSABLE(chunk->mem,
chunk->mem_size());
#endif
for (ulint j = chunk->size;
j--; block++) {