The problem was that when using clang + asan, we do not get a correct value
for the thread stack as some local variables are not allocated at the
normal stack.
It looks like that for example clang 18.1.3, when compiling with
-O2 -fsanitize=addressan it puts local variables and things allocated by
alloca() in other areas than on the stack.
The following code shows the issue
Thread 6 "mariadbd" hit Breakpoint 3, do_handle_one_connection
(connect=0x5080000027b8,
put_in_cache=<optimized out>) at sql/sql_connect.cc:1399
THD *thd;
1399 thd->thread_stack= (char*) &thd;
(gdb) p &thd
(THD **) 0x7fffedee7060
(gdb) p $sp
(void *) 0x7fffef4e7bc0
The address of thd is 24M away from the stack pointer
(gdb) info reg
...
rsp 0x7fffef4e7bc0 0x7fffef4e7bc0
...
r13 0x7fffedee7060 140737185214560
r13 is pointing to the address of the thd. Probably some kind of
"local stack" used by the sanitizer
I have verified this with gdb on a recursive call that calls alloca()
in a loop. In this case all objects was stored in a local heap,
not on the stack.
To solve this issue in a portable way, I have added two functions:
my_get_stack_pointer() returns the address of the current stack pointer.
The code is using asm instructions for intel 32/64 bit, powerpc,
arm 32/64 bit and sparc 32/64 bit.
Supported compilers are gcc, clang and MSVC.
For MSVC 64 bit we are using _AddressOfReturnAddress()
As a fallback for other compilers/arch we use the address of a local
variable.
my_get_stack_bounds() that will return the address of the base stack
and stack size using pthread_attr_getstack() or NtCurrentTed() with
fallback to using the address of a local variable and user provided
stack size.
Server changes are:
- Moving setting of thread_stack to THD::store_globals() using
my_get_stack_bounds().
- Removing setting of thd->thread_stack, except in functions that
allocates a lot on the stack before calling store_globals(). When
using estimates for stack start, we reduce stack_size with
MY_STACK_SAFE_MARGIN (8192) to take into account the stack used
before calling store_globals().
I also added a unittest, stack_allocation-t, to verify the new code.
Reviewed-by: Sergei Golubchik <serg@mariadb.org>
When using the default innodb_log_buffer_size=2m, mariadb-backup --backup
would spend a lot of time re-reading and re-parsing the log. For reads,
it would be beneficial to memory-map the entire ib_logfile0 to the
address space (typically 48 bits or 256 TiB) and read it from there,
both during --backup and --prepare.
We will introduce the Boolean read-only parameter innodb_log_file_mmap
that will be OFF by default on most platforms, to avoid aggressive
read-ahead of the entire ib_logfile0 in when only a tiny portion would be
accessed. On Linux and FreeBSD the default is innodb_log_file_mmap=ON,
because those platforms define a specific mmap(2) option for enabling
such read-ahead and therefore it can be assumed that the default would
be on-demand paging. This parameter will only have impact on the initial
InnoDB startup and recovery. Any writes to the log will use regular I/O,
except when the ib_logfile0 is stored in a specially configured file system
that is backed by persistent memory (Linux "mount -o dax").
We also experimented with allowing writes of the ib_logfile0 via a
memory mapping and decided against it. A fundamental problem would be
unnecessary read-before-write in case of a major page fault, that is,
when a new, not yet cached, virtual memory page in the circular
ib_logfile0 is being written to. There appears to be no way to tell
the operating system that we do not care about the previous contents of
the page, or that the page fault handler should just zero it out.
Many references to HAVE_PMEM have been replaced with references to
HAVE_INNODB_MMAP.
The predicate log_sys.is_pmem() has been replaced with
log_sys.is_mmap() && !log_sys.is_opened().
Memory-mapped regular files differ from MAP_SYNC (PMEM) mappings in the
way that an open file handle to ib_logfile0 will be retained. In both
code paths, log_sys.is_mmap() will hold. Holding a file handle open will
allow log_t::clear_mmap() to disable the interface with fewer operations.
It should be noted that ever since
commit 685d958e38 (MDEV-14425)
most 64-bit Linux platforms on our CI platforms
(s390x a.k.a. IBM System Z being a notable exception) read and write
/dev/shm/*/ib_logfile0 via a memory mapping, pretending that it is
persistent memory (mount -o dax). So, the memory mapping based log
parsing that this change is enabling by default on Linux and FreeBSD
has already been extensively tested on Linux.
::log_mmap(): If a log cannot be opened as PMEM and the desired access
is read-only, try to open a read-only memory mapping.
xtrabackup_copy_mmap_snippet(), xtrabackup_copy_mmap_logfile():
Copy the InnoDB log in mariadb-backup --backup from a memory
mapped file.
Improve detection for DES support in OpenSSL, to allow compilation
against system OpenSSL without DES.
Note that MariaDB needs to be compiled against OpenSSL-like library
that itself has DES support which cmake detected. Positive detection
is indicated with CMake variable HAVE_des 1.
Signed-off-by: Dimitri John Ledkov <dimitri.ledkov@surgut.co.uk>
Apparently, invoking fcntl(fd, F_SETFL, O_DIRECT) will lead to
unexpected behaviour on Linux bcachefs and possibly other file systems,
depending on the operating system version. So, let us avoid doing that,
and instead just attempt to pass the O_DIRECT flag to open(). This should
make us compatible with NetBSD, IBM AIX, as well as Solaris and its
derivatives.
This fix does not change the fact that we had only implemented
innodb_log_file_buffering=OFF on systems where we can determine the
physical block size (typically 512 or 4096 bytes).
Currently, those operating systems are Linux and Microsoft Windows.
HAVE_FCNTL_DIRECT, os_file_set_nocache(): Remove.
OS_FILE_OVERWRITE, OS_FILE_CREATE_PATH: Remove (never used parameters).
os_file_log_buffered(), os_file_log_maybe_unbuffered(): Helper functions.
os_file_create_simple_func(): When applicable, initially attempt to
open files in O_DIRECT mode.
os_file_create_func(): When applicable, initially attempt to
open files in O_DIRECT mode.
For type==OS_LOG_FILE && create_mode != OS_FILE_CREATE
we will first invoke stat(2) on the file name to find out if the size
is compatible with O_DIRECT. If create_mode == OS_FILE_CREATE, we will
invoke fstat(2) on the created log file afterwards, and may close and
reopen the file in O_DIRECT mode if applicable.
create_temp_file(): Support O_DIRECT. This is only used if O_TMPFILE is
available and innodb_disable_sort_file_cache=ON (non-default value).
Notably, that setting never worked on Microsoft Windows.
row_merge_file_create_mode(): Split from row_merge_file_create_low().
Create a temporary file in the specified mode.
Reviewed by: Vladislav Vaintroub
The directio(3C) function on Solaris is supported on NFS and UFS
while the majority of users should be on ZFS, which is a copy-on-write
file system that implements transparent compression and therefore
cannot support unbuffered I/O.
Let us remove the call to directio() and simply treat
innodb_flush_method=O_DIRECT in the same way as the previous
default value innodb_flush_method=fsync on Solaris. Also, let us
remove some dead code around calls to os_file_set_nocache() on
platforms where fcntl(2) is not usable with O_DIRECT.
On IBM AIX, O_DIRECT is not documented for fcntl(2), only for open(2).
Thanks to references from Brad Smith, BSDs use getmntinfo as
a system call for mounted filesystems.
Most BSDs return statfs structures, (and we use OSX's statfs64),
but NetBSD uses a statvfs structure.
Simplify Linux getmntent_r to just use getmntent.
AIX uses getmntent.
An attempt at writing Solaris compatibility with
a small bit of HPUX compatibility was made based on man page
entries only. Fixes welcome.
statvfs structures now use f_bsize for consistency with statfs
Test case adjusted as PATH_MAX is OS defined (e.g. 1023 on AIX)
Fixes: 0ee5cf837e
also fixes:
MDEV-27818: Disk plugin does not show zpool mounted devices
This is because zpool mounted point don't begin with /.
Due to the proliferation of multiple filesystem types since this
was written, we restrict the entries listed in the disks plugin
to excude:
* read only mount points (no point monitoring, and
includes squash, snaps, sysfs, procfs, cgroups...)
* mount points that aren't directories (excludes /etc/hostname and
similar mounts in containers). (getmntent (Linux/AIX) only)
* exclude systems where there is no capacity listed (excludes various
virtual filesystem types).
Reviewer: Sergei Golubchik
As pointed out with MDEV-29308 there are issues with the code as is.
MariaDB is built as C++11 / C99. aligned_alloc() is not guarenteed
to be exposed when building with any mode other than C++17 / C11.
The other *BSD's have their stdlib.h header to expose the function
with C+11 anyway, but the issue exists in the C99 code too, the
build just does not use -Werror. Linux globally defines _GNU_SOURCE
hiding the issue as well.
Table_cache_instance: Define the structure aligned at
the CPU cache line, and remove a pad[] data member.
Krunal Bauskar reported this to improve performance on ARMv8.
aligned_malloc(): Wrapper for the Microsoft _aligned_malloc()
and the ISO/IEC 9899:2011 <stdlib.h> aligned_alloc().
Note: The parameters are in the Microsoft order (size, alignment),
opposite of aligned_alloc(alignment, size).
Note: The standard defines that size must be an integer multiple
of alignment. It is enforced by AddressSanitizer but not by GNU libc
on Linux.
aligned_free(): Wrapper for the Microsoft _aligned_free() and
the standard free().
HAVE_ALIGNED_ALLOC: A new test. Unfortunately, support for
aligned_alloc() may still be missing on some platforms.
We will fall back to posix_memalign() for those cases.
HAVE_MEMALIGN: Remove, along with any use of the nonstandard memalign().
PFS_ALIGNEMENT (sic): Removed; we will use CPU_LEVEL1_DCACHE_LINESIZE.
PFS_ALIGNED: Defined using the C++11 keyword alignas.
buf_pool_t::page_hash_table::create(),
lock_sys_t::hash_table::create():
lock_sys_t::hash_table::resize(): Pad the allocation size to an
integer multiple of the alignment.
Reviewed by: Vladislav Vaintroub
Added checking for support of vfork by a platform where
building being done. Set HAVE_VFORK macros in case vfork()
system call is supported. Use vfork() system call if the
macros HAVE_VFORK is set, else use fork().
This is useful for thing like Item_true and Item_false that we
allocated and initalize once and want to ensure that nothing can
change them
Main changes:
- Memory protection is achived by allocating memory with mmap() and
protect it from write with mprotect()
- init_alloc_root(...,MY_ROOT_USE_MPROTECT) will create a
memroot that one can later use with protect_root() to turn it
read only or turn it back to read-write. All allocations to this
memroot is done with mmap() to ensure page alligned allocations.
- alloc_root() code was rearranged to combine normal and valgrind code.
- init_alloc_root() now changes block size to be power of 2's, to get less
memory fragmentation.
- Changed MEM_ROOT structure to make it smaller. Also renamed
MEM_ROOT m_psi_key to psi_key.
- Moved MY_THREAD_SPECIFIC marker in MEM_ROOT from block size (old hack)
to flags.
- Added global variable my_system_page_size. This is initialized at
startup.
This fixed the MySQL bug# 20338 about misuse of double underscore
prefix __WIN__, which was old MySQL's idea of identifying Windows
Replace it by _WIN32 standard symbol for targeting Windows OS
(both 32 and 64 bit)
Not that connect storage engine is not fixed in this patch (must be
fixed in "upstream" branch)
Systemd has a socket activation feature where a mariadb.socket
definition defines the sockets to listen to, and passes those
file descriptors directly to mariadbd to use when a connection
occurs.
The new functionality is utilized when starting as follows:
systemctl start mariadb.socket
The mariadb.socket definition only needs to contain the network
information, ListenStream= directives, the mariadb.service
definition is still used for service instigation.
When mariadbd is started in this way, the socket, port, bind-address
backlog are all assumed to be self contained in the mariadb.socket
definition and as such the mariadb settings and command line
arguments of these network settings are ignored.
See man systemd.socket for how to limit this to specific ports.
Extra ports, those specified with extra_port in socket activation
mode, are those with a FileDescriptorName=extra. These need
to be in a separate service name like mariadb-extra.socket and
these require a Service={mariadb.service} directive to map to the
original service. Extra ports need systemd v227 or greater
(not RHEL/Centos7 - v219) when FileDescriptorName= was added,
otherwise the extra ports are treated like ordinary ports.
The number of sockets isn't limited when using systemd socket activation
(except by operating system limits on file descriptors and a minimal
amount of memory used per file descriptor). The systemd sockets passed
can include any ownership or permissions, including those the
mariadbd process wouldn't normally have the permission to create.
This implementation is compatible with mariadb.service definitions.
Those services started with:
systemctl start mariadb.service
does actually start the mariadb.service and used all the my.cnf
settings of sockets and ports like it previously did.
`mallinfo` is deprecated since glibc 2.33 and has been replaced by mallinfo2.
The deprecation causes building the server to fail if glibc version is > 2.33.
Check if mallinfo2 exist on the system and use it instead.
