niansa-misc/glibc
1
0
Fork 0
mirror of git://sourceware.org/git/glibc.git synced 2025-03-06 20:58:33 +01:00
Code Issues Projects Releases Packages Wiki Activity

nptl: More useful padding in struct pthread

Browse source 
The previous use of padding within a union made it impossible to
re-use the padding for GLIBC_PRIVATE ABI preservation because
tcbhead_t could use up all of the padding (as was historically the
case on x86-64).  Allocating padding unconditionally addresses this
issue.

Reviewed-by: Adhemerval Zanella  <adhemerval.zanella@linaro.org>
This commit is contained in:
Florian Weimer 2024-12-17 09:20:20 +01:00
parent 5e249192ca
commit 7c22dcda27
1 changed files with 24 additions and 30 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

54
nptl/descr.h
View file

@ -129,41 +129,32 @@ struct priority_protection_data
/* Thread descriptor data structure. */
struct pthread
{
union
{
#if !TLS_DTV_AT_TP
/* This overlaps the TCB as used for TLS without threads (see tls.h). */
tcbhead_t header;
/* This overlaps the TCB as used for TLS without threads (see tls.h). */
tcbhead_t header;
#else
struct
{
/* multiple_threads is enabled either when the process has spawned at
least one thread or when a single-threaded process cancels itself.
This enables additional code to introduce locking before doing some
compare_and_exchange operations and also enable cancellation points.
The concepts of multiple threads and cancellation points ideally
should be separate, since it is not necessary for multiple threads to
have been created for cancellation points to be enabled, as is the
case is when single-threaded process cancels itself.
struct
{
/* multiple_threads is enabled either when the process has spawned at
least one thread or when a single-threaded process cancels itself.
This enables additional code to introduce locking before doing some
compare_and_exchange operations and also enable cancellation points.
The concepts of multiple threads and cancellation points ideally
should be separate, since it is not necessary for multiple threads to
have been created for cancellation points to be enabled, as is the
case is when single-threaded process cancels itself.
Since enabling multiple_threads enables additional code in
cancellation points and compare_and_exchange operations, there is a
potential for an unneeded performance hit when it is enabled in a
single-threaded, self-canceling process. This is OK though, since a
single-threaded process will enable async cancellation only when it
looks to cancel itself and is hence going to end anyway. */
int multiple_threads;
int gscope_flag;
} header;
Since enabling multiple_threads enables additional code in
cancellation points and compare_and_exchange operations, there is a
potential for an unneeded performance hit when it is enabled in a
single-threaded, self-canceling process. This is OK though, since a
single-threaded process will enable async cancellation only when it
looks to cancel itself and is hence going to end anyway. */
int multiple_threads;
int gscope_flag;
} header;
#endif
/* This extra padding has no special purpose, and this structure layout
is private and subject to change without affecting the official ABI.
We just have it here in case it might be convenient for some
implementation-specific instrumentation hack or suchlike. */
void *__padding[24];
};
/* This descriptor's link on the GL (dl_stack_used) or
GL (dl_stack_user) list. */
list_t list;
@ -407,6 +398,9 @@ struct pthread
/* getrandom vDSO per-thread opaque state. */
void *getrandom_buf;
/* Can be used for backports preserving internal TCB layout. */
void *padding[8];
/* rseq area registered with the kernel. Use a custom definition
here to isolate from kernel struct rseq changes. The
implementation of sched_getcpu needs acccess to the cpu_id field;