290 lines
11 KiB
Diff
290 lines
11 KiB
Diff
commit 0c7b002fac12dcb2f53ba83ee56bb3b5d2439447
|
|
Author: Szabolcs Nagy <szabolcs.nagy@arm.com>
|
|
Date: Tue Jun 9 09:57:28 2020 +0100
|
|
|
|
rtld: Add rtld.nns tunable for the number of supported namespaces
|
|
|
|
TLS_STATIC_SURPLUS is 1664 bytes currently which is not enough to
|
|
support DL_NNS (== 16) number of dynamic link namespaces, if we
|
|
assume 192 bytes of TLS are reserved for libc use and 144 bytes
|
|
are reserved for other system libraries that use IE TLS.
|
|
|
|
A new tunable is introduced to control the number of supported
|
|
namespaces and to adjust the surplus static TLS size as follows:
|
|
|
|
surplus_tls = 192 * (rtld.nns-1) + 144 * rtld.nns + 512
|
|
|
|
The default is rtld.nns == 4 and then the surplus TLS size is the
|
|
same as before, so the behaviour is unchanged by default. If an
|
|
application creates more namespaces than the rtld.nns setting
|
|
allows, then it is not guaranteed to work, but the limit is not
|
|
checked. So existing usage will continue to work, but in the
|
|
future if an application creates more than 4 dynamic link
|
|
namespaces then the tunable will need to be set.
|
|
|
|
In this patch DL_NNS is a fixed value and provides a maximum to
|
|
the rtld.nns setting.
|
|
|
|
Static linking used fixed 2048 bytes surplus TLS, this is changed
|
|
so the same contract is used as for dynamic linking. With static
|
|
linking DL_NNS == 1 so rtld.nns tunable is forced to 1, so by
|
|
default the surplus TLS is reduced to 144 + 512 = 656 bytes. This
|
|
change is not expected to cause problems.
|
|
|
|
Tested on aarch64-linux-gnu and x86_64-linux-gnu.
|
|
|
|
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
|
|
|
|
Conflicts:
|
|
elf/dl-tls.c
|
|
(Different per-namespace TLS reservation defaults before
|
|
this backport.)
|
|
|
|
diff --git a/csu/libc-tls.c b/csu/libc-tls.c
|
|
index 28a79441cde379f7..08ed2b988b58ac6c 100644
|
|
--- a/csu/libc-tls.c
|
|
+++ b/csu/libc-tls.c
|
|
@@ -52,13 +52,16 @@ bool _dl_tls_dtv_gaps;
|
|
struct dtv_slotinfo_list *_dl_tls_dtv_slotinfo_list;
|
|
/* Number of modules in the static TLS block. */
|
|
size_t _dl_tls_static_nelem;
|
|
-/* Size of the static TLS block. Giving this initialized value
|
|
- preallocates some surplus bytes in the static TLS area. */
|
|
-size_t _dl_tls_static_size = 2048;
|
|
+/* Size of the static TLS block. */
|
|
+size_t _dl_tls_static_size;
|
|
/* Size actually allocated in the static TLS block. */
|
|
size_t _dl_tls_static_used;
|
|
/* Alignment requirement of the static TLS block. */
|
|
size_t _dl_tls_static_align;
|
|
+/* Size of surplus space in the static TLS area for dynamically
|
|
+ loaded modules with IE-model TLS or for TLSDESC optimization.
|
|
+ See comments in elf/dl-tls.c where it is initialized. */
|
|
+size_t _dl_tls_static_surplus;
|
|
|
|
/* Generation counter for the dtv. */
|
|
size_t _dl_tls_generation;
|
|
@@ -87,10 +90,8 @@ init_slotinfo (void)
|
|
static void
|
|
init_static_tls (size_t memsz, size_t align)
|
|
{
|
|
- /* That is the size of the TLS memory for this object. The initialized
|
|
- value of _dl_tls_static_size is provided by dl-open.c to request some
|
|
- surplus that permits dynamic loading of modules with IE-model TLS. */
|
|
- GL(dl_tls_static_size) = roundup (memsz + GL(dl_tls_static_size),
|
|
+ /* That is the size of the TLS memory for this object. */
|
|
+ GL(dl_tls_static_size) = roundup (memsz + GLRO(dl_tls_static_surplus),
|
|
TLS_TCB_ALIGN);
|
|
#if TLS_TCB_AT_TP
|
|
GL(dl_tls_static_size) += TLS_TCB_SIZE;
|
|
@@ -131,25 +132,24 @@ __libc_setup_tls (void)
|
|
break;
|
|
}
|
|
|
|
+ /* Calculate the size of the static TLS surplus. */
|
|
+ _dl_tls_static_surplus_init ();
|
|
+
|
|
/* We have to set up the TCB block which also (possibly) contains
|
|
'errno'. Therefore we avoid 'malloc' which might touch 'errno'.
|
|
Instead we use 'sbrk' which would only uses 'errno' if it fails.
|
|
In this case we are right away out of memory and the user gets
|
|
- what she/he deserves.
|
|
-
|
|
- The initialized value of _dl_tls_static_size is provided by dl-open.c
|
|
- to request some surplus that permits dynamic loading of modules with
|
|
- IE-model TLS. */
|
|
+ what she/he deserves. */
|
|
#if TLS_TCB_AT_TP
|
|
/* Align the TCB offset to the maximum alignment, as
|
|
_dl_allocate_tls_storage (in elf/dl-tls.c) does using __libc_memalign
|
|
and dl_tls_static_align. */
|
|
- tcb_offset = roundup (memsz + GL(dl_tls_static_size), max_align);
|
|
+ tcb_offset = roundup (memsz + GLRO(dl_tls_static_surplus), max_align);
|
|
tlsblock = __sbrk (tcb_offset + TLS_INIT_TCB_SIZE + max_align);
|
|
#elif TLS_DTV_AT_TP
|
|
tcb_offset = roundup (TLS_INIT_TCB_SIZE, align ?: 1);
|
|
tlsblock = __sbrk (tcb_offset + memsz + max_align
|
|
- + TLS_PRE_TCB_SIZE + GL(dl_tls_static_size));
|
|
+ + TLS_PRE_TCB_SIZE + GLRO(dl_tls_static_surplus));
|
|
tlsblock += TLS_PRE_TCB_SIZE;
|
|
#else
|
|
/* In case a model with a different layout for the TCB and DTV
|
|
diff --git a/elf/dl-tls.c b/elf/dl-tls.c
|
|
index a2def280b7096960..ef57a21391bb36fa 100644
|
|
--- a/elf/dl-tls.c
|
|
+++ b/elf/dl-tls.c
|
|
@@ -29,10 +29,54 @@
|
|
#include <dl-tls.h>
|
|
#include <ldsodefs.h>
|
|
|
|
-/* Amount of excess space to allocate in the static TLS area
|
|
- to allow dynamic loading of modules defining IE-model TLS data. */
|
|
-#define TLS_STATIC_SURPLUS 64 + DL_NNS * 100
|
|
+#define TUNABLE_NAMESPACE rtld
|
|
+#include <dl-tunables.h>
|
|
+
|
|
+/* Surplus static TLS, GLRO(dl_tls_static_surplus), is used for
|
|
+
|
|
+ - IE TLS in libc.so for all dlmopen namespaces except in the initial
|
|
+ one where libc.so is not loaded dynamically but at startup time,
|
|
+ - IE TLS in other libraries which may be dynamically loaded even in the
|
|
+ initial namespace,
|
|
+ - and optionally for optimizing dynamic TLS access.
|
|
+
|
|
+ The maximum number of namespaces is DL_NNS, but to support that many
|
|
+ namespaces correctly the static TLS allocation should be significantly
|
|
+ increased, which may cause problems with small thread stacks due to the
|
|
+ way static TLS is accounted (bug 11787).
|
|
+
|
|
+ So there is a rtld.nns tunable limit on the number of supported namespaces
|
|
+ that affects the size of the static TLS and by default it's small enough
|
|
+ not to cause problems with existing applications. The limit is not
|
|
+ enforced or checked: it is the user's responsibility to increase rtld.nns
|
|
+ if more dlmopen namespaces are used. */
|
|
+
|
|
+/* Size of initial-exec TLS in libc.so. */
|
|
+#define LIBC_IE_TLS 192
|
|
+/* Size of initial-exec TLS in libraries other than libc.so.
|
|
+ This should be large enough to cover runtime libraries of the
|
|
+ compiler such as libgomp and libraries in libc other than libc.so. */
|
|
+#define OTHER_IE_TLS 144
|
|
+/* Size of additional surplus TLS, placeholder for TLS optimizations. */
|
|
+#define OPT_SURPLUS_TLS 512
|
|
|
|
+void
|
|
+_dl_tls_static_surplus_init (void)
|
|
+{
|
|
+ size_t nns;
|
|
+
|
|
+#if HAVE_TUNABLES
|
|
+ nns = TUNABLE_GET (nns, size_t, NULL);
|
|
+#else
|
|
+ /* Default values of the tunables. */
|
|
+ nns = 4;
|
|
+#endif
|
|
+ if (nns > DL_NNS)
|
|
+ nns = DL_NNS;
|
|
+ GLRO(dl_tls_static_surplus) = ((nns - 1) * LIBC_IE_TLS
|
|
+ + nns * OTHER_IE_TLS
|
|
+ + OPT_SURPLUS_TLS);
|
|
+}
|
|
|
|
/* Out-of-memory handler. */
|
|
static void
|
|
@@ -218,7 +262,8 @@ _dl_determine_tlsoffset (void)
|
|
}
|
|
|
|
GL(dl_tls_static_used) = offset;
|
|
- GL(dl_tls_static_size) = (roundup (offset + TLS_STATIC_SURPLUS, max_align)
|
|
+ GL(dl_tls_static_size) = (roundup (offset + GLRO(dl_tls_static_surplus),
|
|
+ max_align)
|
|
+ TLS_TCB_SIZE);
|
|
#elif TLS_DTV_AT_TP
|
|
/* The TLS blocks start right after the TCB. */
|
|
@@ -262,7 +307,7 @@ _dl_determine_tlsoffset (void)
|
|
}
|
|
|
|
GL(dl_tls_static_used) = offset;
|
|
- GL(dl_tls_static_size) = roundup (offset + TLS_STATIC_SURPLUS,
|
|
+ GL(dl_tls_static_size) = roundup (offset + GLRO(dl_tls_static_surplus),
|
|
TLS_TCB_ALIGN);
|
|
#else
|
|
# error "Either TLS_TCB_AT_TP or TLS_DTV_AT_TP must be defined"
|
|
diff --git a/elf/dl-tunables.list b/elf/dl-tunables.list
|
|
index b7cc79f8bfe0a7c6..7337fb85062c91a7 100644
|
|
--- a/elf/dl-tunables.list
|
|
+++ b/elf/dl-tunables.list
|
|
@@ -126,4 +126,13 @@ glibc {
|
|
default: 3
|
|
}
|
|
}
|
|
+
|
|
+ rtld {
|
|
+ nns {
|
|
+ type: SIZE_T
|
|
+ minval: 1
|
|
+ maxval: 16
|
|
+ default: 4
|
|
+ }
|
|
+ }
|
|
}
|
|
diff --git a/elf/rtld.c b/elf/rtld.c
|
|
index 772aff5160359b7b..a440741f4c1b3c91 100644
|
|
--- a/elf/rtld.c
|
|
+++ b/elf/rtld.c
|
|
@@ -776,6 +776,9 @@ init_tls (void)
|
|
}
|
|
assert (i == GL(dl_tls_max_dtv_idx));
|
|
|
|
+ /* Calculate the size of the static TLS surplus. */
|
|
+ _dl_tls_static_surplus_init ();
|
|
+
|
|
/* Compute the TLS offsets for the various blocks. */
|
|
_dl_determine_tlsoffset ();
|
|
|
|
diff --git a/manual/tunables.texi b/manual/tunables.texi
|
|
index 55d5dfb14db4dfb8..e092b8e81a18d739 100644
|
|
--- a/manual/tunables.texi
|
|
+++ b/manual/tunables.texi
|
|
@@ -31,6 +31,7 @@ their own namespace.
|
|
@menu
|
|
* Tunable names:: The structure of a tunable name
|
|
* Memory Allocation Tunables:: Tunables in the memory allocation subsystem
|
|
+* Dynamic Linking Tunables:: Tunables in the dynamic linking subsystem
|
|
* Elision Tunables:: Tunables in elision subsystem
|
|
* Hardware Capability Tunables:: Tunables that modify the hardware
|
|
capabilities seen by @theglibc{}
|
|
@@ -225,6 +226,26 @@ pointer, so add 4 on 32-bit systems or 8 on 64-bit systems to the size
|
|
passed to @code{malloc} for the largest bin size to enable.
|
|
@end deftp
|
|
|
|
+@node Dynamic Linking Tunables
|
|
+@section Dynamic Linking Tunables
|
|
+@cindex dynamic linking tunables
|
|
+@cindex rtld tunables
|
|
+
|
|
+@deftp {Tunable namespace} glibc.rtld
|
|
+Dynamic linker behavior can be modified by setting the
|
|
+following tunables in the @code{rtld} namespace:
|
|
+@end deftp
|
|
+
|
|
+@deftp Tunable glibc.rtld.nns
|
|
+Sets the number of supported dynamic link namespaces (see @code{dlmopen}).
|
|
+Currently this limit can be set between 1 and 16 inclusive, the default is 4.
|
|
+Each link namespace consumes some memory in all thread, and thus raising the
|
|
+limit will increase the amount of memory each thread uses. Raising the limit
|
|
+is useful when your application uses more than 4 dynamic linker audit modules
|
|
+e.g. @env{LD_AUDIT}, or will use more than 4 dynamic link namespaces as created
|
|
+by @code{dlmopen} with an lmid argument of @code{LM_ID_NEWLM}.
|
|
+@end deftp
|
|
+
|
|
@node Elision Tunables
|
|
@section Elision Tunables
|
|
@cindex elision tunables
|
|
diff --git a/sysdeps/generic/ldsodefs.h b/sysdeps/generic/ldsodefs.h
|
|
index ccec08929e4ad4e7..e54105848c3cb7d1 100644
|
|
--- a/sysdeps/generic/ldsodefs.h
|
|
+++ b/sysdeps/generic/ldsodefs.h
|
|
@@ -582,6 +582,11 @@ struct rtld_global_ro
|
|
binaries, don't honor for PIEs). */
|
|
EXTERN ElfW(Addr) _dl_use_load_bias;
|
|
|
|
+ /* Size of surplus space in the static TLS area for dynamically
|
|
+ loaded modules with IE-model TLS or for TLSDESC optimization.
|
|
+ See comments in elf/dl-tls.c where it is initialized. */
|
|
+ EXTERN size_t _dl_tls_static_surplus;
|
|
+
|
|
/* Name of the shared object to be profiled (if any). */
|
|
EXTERN const char *_dl_profile;
|
|
/* Filename of the output file. */
|
|
@@ -1099,6 +1104,9 @@ extern size_t _dl_count_modids (void) attribute_hidden;
|
|
/* Calculate offset of the TLS blocks in the static TLS block. */
|
|
extern void _dl_determine_tlsoffset (void) attribute_hidden;
|
|
|
|
+/* Calculate the size of the static TLS surplus. */
|
|
+void _dl_tls_static_surplus_init (void) attribute_hidden;
|
|
+
|
|
#ifndef SHARED
|
|
/* Set up the TCB for statically linked applications. This is called
|
|
early during startup because we always use TLS (for errno and the
|