| Message ID | 20200326155633.18236-6-mathieu.desnoyers@efficios.com |
|---|---|
| State | New |
| Headers | show |
| Series | Restartable Sequences enablement | expand |
* Mathieu Desnoyers via Libc-alpha: > +* Support for automatically registering threads with the Linux rseq(2) > + system call has been added. This system call is implemented starting > + from Linux 4.18. The Restartable Sequences ABI accelerates user-space > + operations on per-cpu data. It allows user-space to perform updates > + on per-cpu data without requiring heavy-weight atomic operations. > + Automatically registering threads allows all libraries, including libc, > + to make immediate use of the rseq(2) support by using the documented ABI. > + See 'man 2 rseq' for the details of the ABI shared between libc and the > + kernel. This should refer documentation in the glibc manual. (It is currently a glibc project requirement to add documentation for new Linux interfaces, something that I do not necessarily agree with.) > > Deprecated and removed features, and other changes affecting compatibility: > > diff --git a/elf/libc_early_init.c b/elf/libc_early_init.c > index 1ac66d895d..30466afea0 100644 > --- a/elf/libc_early_init.c > +++ b/elf/libc_early_init.c > @@ -18,10 +18,13 @@ > > #include <ctype.h> > #include <libc-early-init.h> > +#include <rseq-internal.h> > > void > __libc_early_init (void) > { > /* Initialize ctype data. */ > __ctype_init (); > + /* Register rseq ABI to the kernel. */ > + (void) rseq_register_current_thread (); > } The cast to void should be removed (see below the comment about the return type. > diff --git a/manual/threads.texi b/manual/threads.texi > index 0858ef8f92..59f634e432 100644 > --- a/manual/threads.texi > +++ b/manual/threads.texi > @@ -9,8 +9,10 @@ This chapter describes functions used for managing threads. > POSIX threads. > > @menu > -* ISO C Threads:: Threads based on the ISO C specification. > -* POSIX Threads:: Threads based on the POSIX specification. > +* ISO C Threads:: Threads based on the ISO C specification. > +* POSIX Threads:: Threads based on the POSIX specification. > +* Restartable Sequences:: Linux-specific Restartable Sequences > + integration. > @end menu > > > @@ -881,3 +883,27 @@ Behaves like @code{pthread_timedjoin_np} except that the absolute time in > @c pthread_spin_unlock > @c pthread_testcancel > @c pthread_yield > + > +@node Restartable Sequences > +@section Restartable Sequences > +@cindex rseq > + > +This section describes @theglibc{} Restartable Sequences integration. > + > +@deftypevar {struct rseq} __rseq_abi > +@standards{GNU, sys/rseq.h} > +@Theglibc{} implements a @code{__rseq_abi} TLS symbol to interact with the > +Restartable Sequences system call (Linux-specific). The layout of this > +structure is defined by the Linux kernel @file{linux/rseq.h} UAPI. > +Registration of each thread's @code{__rseq_abi} is performed by > +@theglibc{} at libc initialization and pthread creation. > +@end deftypevar > + > +@deftypevr Macro int RSEQ_SIG > +@standards{GNU, sys/rseq.h} > +Each supported architecture provide a @code{RSEQ_SIG} macro in > +@file{sys/rseq.h} which contains a signature. That signature is expected to be > +present in the code before each Restartable Sequences abort handler. Failure > +to provide the expected signature may terminate the process with a Segmentation > +fault. > +@end deftypevr This should say Linux, not GNU as the standards source, given that the interface is not added to the GNU ABI. Is this sufficient documentation of this feature? What do others think about this? > diff --git a/misc/rseq-internal.h b/misc/rseq-internal.h > new file mode 100644 > index 0000000000..d564cf1bc3 > --- /dev/null > +++ b/misc/rseq-internal.h > @@ -0,0 +1,33 @@ > +static inline int > +rseq_register_current_thread (void) > +{ > + return -1; > +} Nothing checks the return value of this function as far as I can see, so it can return void. > +static inline int > +rseq_unregister_current_thread (void) > +{ > + return -1; > +} This function is unused. This also applies to the full version. I believe we switched to implicit deregistration on thread exit, so I think you can just remove it. > diff --git a/nptl/pthread_create.c b/nptl/pthread_create.c > index afd379e89a..1ff248042e 100644 > --- a/nptl/pthread_create.c > +++ b/nptl/pthread_create.c > @@ -33,6 +33,7 @@ > #include <default-sched.h> > #include <futex-internal.h> > #include <tls-setup.h> > +#include <rseq-internal.h> > #include "libioP.h" > > #include <shlib-compat.h> > @@ -384,6 +385,9 @@ START_THREAD_DEFN > /* Initialize pointers to locale data. */ > __ctype_init (); > > + /* Register rseq TLS to the kernel. */ > + (void) rseq_register_current_thread (); > + The cast can go. > #ifndef __ASSUME_SET_ROBUST_LIST > if (__set_robust_list_avail >= 0) > #endif > @@ -578,6 +582,14 @@ START_THREAD_DEFN > process is really dead since 'clone' got passed the CLONE_CHILD_CLEARTID > flag. The 'tid' field in the TCB will be set to zero. > > + rseq TLS is still registered at this point. Rely on implicit unregistration > + performed by the kernel on thread teardown. This is not a problem because the > + rseq TLS lives on the stack, and the stack outlives the thread. If TCB > + allocation is ever changed, additional steps may be required, such as > + performing explicit rseq unregistration before reclaiming the rseq TLS area > + memory. It is NOT sufficient to block signals because the kernel may write > + to the rseq area even without signals. > + > The exit code is zero since in case all threads exit by calling > 'pthread_exit' the exit status must be 0 (zero). */ > __exit_thread (); Some of these lines are too long. Also two spaces after . at the end of sentences. > diff --git a/sysdeps/unix/sysv/linux/rseq-internal.h b/sysdeps/unix/sysv/linux/rseq-internal.h > new file mode 100644 > index 0000000000..5f7f02f1ec > --- /dev/null > +++ b/sysdeps/unix/sysv/linux/rseq-internal.h > +static inline int > +rseq_register_current_thread (void) > +{ > + int rc, ret = 0; > + > + if (__rseq_abi.cpu_id == RSEQ_CPU_ID_REGISTRATION_FAILED) > + return -1; > + rc = INTERNAL_SYSCALL_CALL (rseq, &__rseq_abi, sizeof (struct rseq), > + 0, RSEQ_SIG); > + if (!rc) > + goto end; > + if (INTERNAL_SYSCALL_ERRNO (rc) != EBUSY) > + __rseq_abi.cpu_id = RSEQ_CPU_ID_REGISTRATION_FAILED; > + ret = -1; > +end: > + return ret; > +} This does not seem to use INTERNAL_SYSCALL_CALL correctly. I think you need to use INTERNAL_SYSCALL_ERROR_P on the result to check for an error, and only then use INTERNAL_SYSCALL_ERRNO to extract the error code. > diff --git a/sysdeps/unix/sysv/linux/rseq-sym.c b/sysdeps/unix/sysv/linux/rseq-sym.c > new file mode 100644 > index 0000000000..0e33fab278 > --- /dev/null > +++ b/sysdeps/unix/sysv/linux/rseq-sym.c > +#include <sys/syscall.h> > +#include <stdint.h> > +#include <kernel-features.h> > +#include <sys/rseq.h> > + > +__thread struct rseq __rseq_abi = { > + .cpu_id = RSEQ_CPU_ID_UNINITIALIZED, > +}; { should go onto its own line. I'd also add attribute_tls_model_ie, also it's implied by the declaration in the header. > diff --git a/sysdeps/unix/sysv/linux/sys/rseq.h b/sysdeps/unix/sysv/linux/sys/rseq.h > new file mode 100644 > index 0000000000..503dce4cac > --- /dev/null > +++ b/sysdeps/unix/sysv/linux/sys/rseq.h > @@ -0,0 +1,186 @@ I think there is some value in making this header compatible with inclusion from the assembler (including constants for the relevant struct offsets), but that can be a later change. > +#ifndef _SYS_RSEQ_H > +#define _SYS_RSEQ_H 1 > + > +/* Architecture-specific rseq signature. */ > +#include <bits/rseq.h> Maybe add a newline between the above and the following, to make clear the comment only applies to the first #include. > +#include <stdint.h> > +#include <sys/cdefs.h> > + > +#ifdef __has_include > +# if __has_include ("linux/rseq.h") > +# define __GLIBC_HAVE_KERNEL_RSEQ > +# endif > +#else > +# include <linux/version.h> > +# if LINUX_VERSION_CODE >= KERNEL_VERSION (4, 18, 0) > +# define __GLIBC_HAVE_KERNEL_RSEQ > +# endif > +#endif > + > +#ifdef __GLIBC_HAVE_KERNEL_RSEQ > +/* We use the structures declarations from the kernel headers. */ > +# include <linux/rseq.h> > +#else > +/* We use a copy of the include/uapi/linux/rseq.h kernel header. */ > + > +#include <asm/byteorder.h> > + > +enum rseq_cpu_id_state > + { > + RSEQ_CPU_ID_UNINITIALIZED = -1, > + RSEQ_CPU_ID_REGISTRATION_FAILED = -2, > + }; > + > +enum rseq_flags > + { > + RSEQ_FLAG_UNREGISTER = (1 << 0), > + }; > + > +enum rseq_cs_flags_bit > + { > + RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT = 0, > + RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT = 1, > + RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT = 2, > + }; > + > +enum rseq_cs_flags > + { > + RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT = > + (1U << RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT), > + RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL = > + (1U << RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT), > + RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE = > + (1U << RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT), > + }; > + > +/* struct rseq_cs is aligned on 4 * 8 bytes to ensure it is always > + contained within a single cache-line. It is usually declared as > + link-time constant data. */ > +struct rseq_cs > + { > + /* Version of this structure. */ > + uint32_t version; > + /* enum rseq_cs_flags. */ > + uint32_t flags; > + uint64_t start_ip; > + /* Offset from start_ip. */ > + uint64_t post_commit_offset; > + uint64_t abort_ip; > +} __attribute__((aligned(4 * sizeof(uint64_t)))); The comment is wrong. 32-byte alignment does not put struct rseq_cs on its own cache line on many (most?) CPUs. Not using the constant 32 looks like unnecessary obfuscation to me. I still think we should avoid the alignment. The _ip fields should perhaps be _pc (IP is more or less specific to x86). { and } are not aligned. Please do not forget to add spaces before opening parentheses, and two spaces after the . and the end of sentences. The opening { should always be on its own line. (This also applies to the definition of struct rseq below.) > + > +/* struct rseq is aligned on 4 * 8 bytes to ensure it is always > + contained within a single cache-line. > + > + A single struct rseq per thread is allowed. */ > +struct rseq > + { > + /* Restartable sequences cpu_id_start field. Updated by the > + kernel. Read by user-space with single-copy atomicity > + semantics. This field should only be read by the thread which > + registered this data structure. Aligned on 32-bit. Always What does “Aligned on 32-bit” mean in this context? Do you mean to reference 32-*byte* alignment here? > + /* Restartable sequences rseq_cs field. > + > + Contains NULL when no critical section is active for the current > + thread, or holds a pointer to the currently active struct rseq_cs. > + > + Updated by user-space, which sets the address of the currently > + active rseq_cs at the beginning of assembly instruction sequence > + block, and set to NULL by the kernel when it restarts an assembly > + instruction sequence block, as well as when the kernel detects that > + it is preempting or delivering a signal outside of the range > + targeted by the rseq_cs. Also needs to be set to NULL by user-space > + before reclaiming memory that contains the targeted struct rseq_cs. > + > + Read and set by the kernel. Set by user-space with single-copy > + atomicity semantics. This field should only be updated by the > + thread which registered this data structure. Aligned on 64-bit. */ > + union { > + uint64_t ptr64; > +#ifdef __LP64__ > + uint64_t ptr; > +#else > + struct { > +#if (defined(__BYTE_ORDER) && (__BYTE_ORDER == __BIG_ENDIAN)) || defined(__BIG_ENDIAN) > + uint32_t padding; /* Initialized to zero. */ > + uint32_t ptr32; > +#else /* LITTLE */ > + uint32_t ptr32; > + uint32_t padding; /* Initialized to zero. */ > +#endif /* ENDIAN */ > + } ptr; > +#endif > + } rseq_cs; Are these conditionals correct for x32? Shouldn't there be a member of type const struct rseq_cs * somewhere? > diff --git a/sysdeps/unix/sysv/linux/x86/bits/rseq.h b/sysdeps/unix/sysv/linux/x86/bits/rseq.h > new file mode 100644 > index 0000000000..75f52d9788 > --- /dev/null > +++ b/sysdeps/unix/sysv/linux/x86/bits/rseq.h > @@ -0,0 +1,30 @@ > +/* Restartable Sequences Linux x86 architecture header. > + Copyright (C) 2019-2020 Free Software Foundation, Inc. Please make sure that none of the new files reference the 2019 year. It should be 2020, per GNU policy. The patch needs some rebasing on top of current master.
* Mathieu Desnoyers via Libc-alpha: > diff --git a/elf/libc_early_init.c b/elf/libc_early_init.c > index 1ac66d895d..30466afea0 100644 > --- a/elf/libc_early_init.c > +++ b/elf/libc_early_init.c > @@ -18,10 +18,13 @@ > > #include <ctype.h> > #include <libc-early-init.h> > +#include <rseq-internal.h> > > void > __libc_early_init (void) > { > /* Initialize ctype data. */ > __ctype_init (); > + /* Register rseq ABI to the kernel. */ > + (void) rseq_register_current_thread (); > } I think the registration must be restricted to the primary namespace. Otherwise, LD_AUDIT will register the area to the secondary libc (in the audit module), not the primary libc for the entire process. I think the easiest way to implement this for now is a flag argument for __libc_early_init (as the upstream __libc_multiple_libcs is not entirely accurate). I will submit a patch.
----- On Apr 27, 2020, at 5:11 AM, Florian Weimer fw@deneb.enyo.de wrote: > * Mathieu Desnoyers via Libc-alpha: > >> +* Support for automatically registering threads with the Linux rseq(2) >> + system call has been added. This system call is implemented starting >> + from Linux 4.18. The Restartable Sequences ABI accelerates user-space >> + operations on per-cpu data. It allows user-space to perform updates >> + on per-cpu data without requiring heavy-weight atomic operations. >> + Automatically registering threads allows all libraries, including libc, >> + to make immediate use of the rseq(2) support by using the documented ABI. >> + See 'man 2 rseq' for the details of the ABI shared between libc and the >> + kernel. > > This should refer documentation in the glibc manual. > > (It is currently a glibc project requirement to add documentation for > new Linux interfaces, something that I do not necessarily agree with.) A related issue here is that editing of the rseq(2) man pages has been stalled since March 2019. I have been waiting for reply from Michael Kerrisk, and I suspect he might have be side-tracked by other projects. I just bumped that thread. Ref.: https://lore.kernel.org/r/211707091.921.1551722548347.JavaMail.zimbra@efficios.com So as of today, "man 2 rseq" does not exist in the kernel man pages, so I suggest we remove that sentence. Would the following change be OK with you ? - See 'man 2 rseq' for the details of the ABI shared between libc and the - kernel. + The GNU C Library manual has details on integration of Restartable + Sequences. > >> >> Deprecated and removed features, and other changes affecting compatibility: >> >> diff --git a/elf/libc_early_init.c b/elf/libc_early_init.c >> index 1ac66d895d..30466afea0 100644 >> --- a/elf/libc_early_init.c >> +++ b/elf/libc_early_init.c >> @@ -18,10 +18,13 @@ >> >> #include <ctype.h> >> #include <libc-early-init.h> >> +#include <rseq-internal.h> >> >> void >> __libc_early_init (void) >> { >> /* Initialize ctype data. */ >> __ctype_init (); >> + /* Register rseq ABI to the kernel. */ >> + (void) rseq_register_current_thread (); >> } > > The cast to void should be removed (see below the comment about the > return type. OK > >> diff --git a/manual/threads.texi b/manual/threads.texi >> index 0858ef8f92..59f634e432 100644 >> --- a/manual/threads.texi >> +++ b/manual/threads.texi >> @@ -9,8 +9,10 @@ This chapter describes functions used for managing threads. >> POSIX threads. >> >> @menu >> -* ISO C Threads:: Threads based on the ISO C specification. >> -* POSIX Threads:: Threads based on the POSIX specification. >> +* ISO C Threads:: Threads based on the ISO C specification. >> +* POSIX Threads:: Threads based on the POSIX specification. >> +* Restartable Sequences:: Linux-specific Restartable Sequences >> + integration. >> @end menu >> >> >> @@ -881,3 +883,27 @@ Behaves like @code{pthread_timedjoin_np} except that the >> absolute time in >> @c pthread_spin_unlock >> @c pthread_testcancel >> @c pthread_yield >> + >> +@node Restartable Sequences >> +@section Restartable Sequences >> +@cindex rseq >> + >> +This section describes @theglibc{} Restartable Sequences integration. >> + >> +@deftypevar {struct rseq} __rseq_abi >> +@standards{GNU, sys/rseq.h} >> +@Theglibc{} implements a @code{__rseq_abi} TLS symbol to interact with the >> +Restartable Sequences system call (Linux-specific). The layout of this >> +structure is defined by the Linux kernel @file{linux/rseq.h} UAPI. >> +Registration of each thread's @code{__rseq_abi} is performed by >> +@theglibc{} at libc initialization and pthread creation. >> +@end deftypevar >> + >> +@deftypevr Macro int RSEQ_SIG >> +@standards{GNU, sys/rseq.h} >> +Each supported architecture provide a @code{RSEQ_SIG} macro in >> +@file{sys/rseq.h} which contains a signature. That signature is expected to be >> +present in the code before each Restartable Sequences abort handler. Failure >> +to provide the expected signature may terminate the process with a Segmentation >> +fault. >> +@end deftypevr > > This should say Linux, not GNU as the standards source, given that the > interface is not added to the GNU ABI. Good point, done. > > Is this sufficient documentation of this feature? What do others > think about this? I'll leave this question to others. > >> diff --git a/misc/rseq-internal.h b/misc/rseq-internal.h >> new file mode 100644 >> index 0000000000..d564cf1bc3 >> --- /dev/null >> +++ b/misc/rseq-internal.h >> @@ -0,0 +1,33 @@ > >> +static inline int >> +rseq_register_current_thread (void) >> +{ >> + return -1; >> +} > > Nothing checks the return value of this function as far as I can see, > so it can return void. OK > >> +static inline int >> +rseq_unregister_current_thread (void) >> +{ >> + return -1; >> +} > > This function is unused. This also applies to the full version. I > believe we switched to implicit deregistration on thread exit, so I > think you can just remove it. OK > >> diff --git a/nptl/pthread_create.c b/nptl/pthread_create.c >> index afd379e89a..1ff248042e 100644 >> --- a/nptl/pthread_create.c >> +++ b/nptl/pthread_create.c >> @@ -33,6 +33,7 @@ >> #include <default-sched.h> >> #include <futex-internal.h> >> #include <tls-setup.h> >> +#include <rseq-internal.h> >> #include "libioP.h" >> >> #include <shlib-compat.h> >> @@ -384,6 +385,9 @@ START_THREAD_DEFN >> /* Initialize pointers to locale data. */ >> __ctype_init (); >> >> + /* Register rseq TLS to the kernel. */ >> + (void) rseq_register_current_thread (); >> + > > The cast can go. OK > >> #ifndef __ASSUME_SET_ROBUST_LIST >> if (__set_robust_list_avail >= 0) >> #endif >> @@ -578,6 +582,14 @@ START_THREAD_DEFN >> process is really dead since 'clone' got passed the CLONE_CHILD_CLEARTID >> flag. The 'tid' field in the TCB will be set to zero. >> >> + rseq TLS is still registered at this point. Rely on implicit >> unregistration >> + performed by the kernel on thread teardown. This is not a problem because >> the >> + rseq TLS lives on the stack, and the stack outlives the thread. If TCB >> + allocation is ever changed, additional steps may be required, such as >> + performing explicit rseq unregistration before reclaiming the rseq TLS >> area >> + memory. It is NOT sufficient to block signals because the kernel may write >> + to the rseq area even without signals. >> + >> The exit code is zero since in case all threads exit by calling >> 'pthread_exit' the exit status must be 0 (zero). */ >> __exit_thread (); > > Some of these lines are too long. Also two spaces after . at the end > of sentences. OK > >> diff --git a/sysdeps/unix/sysv/linux/rseq-internal.h >> b/sysdeps/unix/sysv/linux/rseq-internal.h >> new file mode 100644 >> index 0000000000..5f7f02f1ec >> --- /dev/null >> +++ b/sysdeps/unix/sysv/linux/rseq-internal.h > >> +static inline int >> +rseq_register_current_thread (void) >> +{ >> + int rc, ret = 0; >> + >> + if (__rseq_abi.cpu_id == RSEQ_CPU_ID_REGISTRATION_FAILED) >> + return -1; >> + rc = INTERNAL_SYSCALL_CALL (rseq, &__rseq_abi, sizeof (struct rseq), >> + 0, RSEQ_SIG); >> + if (!rc) >> + goto end; >> + if (INTERNAL_SYSCALL_ERRNO (rc) != EBUSY) >> + __rseq_abi.cpu_id = RSEQ_CPU_ID_REGISTRATION_FAILED; >> + ret = -1; >> +end: >> + return ret; >> +} > > This does not seem to use INTERNAL_SYSCALL_CALL correctly. I think > you need to use INTERNAL_SYSCALL_ERROR_P on the result to check for an > error, and only then use INTERNAL_SYSCALL_ERRNO to extract the error > code. OK > >> diff --git a/sysdeps/unix/sysv/linux/rseq-sym.c >> b/sysdeps/unix/sysv/linux/rseq-sym.c >> new file mode 100644 >> index 0000000000..0e33fab278 >> --- /dev/null >> +++ b/sysdeps/unix/sysv/linux/rseq-sym.c > >> +#include <sys/syscall.h> >> +#include <stdint.h> >> +#include <kernel-features.h> >> +#include <sys/rseq.h> >> + >> +__thread struct rseq __rseq_abi = { >> + .cpu_id = RSEQ_CPU_ID_UNINITIALIZED, >> +}; > > { should go onto its own line. OK > I'd also add attribute_tls_model_ie, > also it's implied by the declaration in the header. This contradicts feedback I received from Szabolcs Nagy in September 2019: https://public-inbox.org/libc-alpha/c58d4d6e-f22a-f5d9-e23a-5bd72cec1a86@arm.com/ "note that libpthread.so is built with -ftls-model=initial-exec (and if it wasn't then you'd want to put the attribute on the declaration in the internal header file, not on the definition, so the actual tls accesses generate the right code)" In the context of his feedback, __rseq_abi was defined within nptl/pthread_create.c. It is now defined in sysdeps/unix/sysv/linux/rseq-sym.c, which is built into the csu which AFAIU ends up in libc.so. His comment still applies though, because libc.so is also built with -ftls-model=initial-exec. So should I apply the "initial-exec" TLS model only to the __rseq_abi declaration, or is it preferred to apply it to both the declaration and the definition ? > >> diff --git a/sysdeps/unix/sysv/linux/sys/rseq.h >> b/sysdeps/unix/sysv/linux/sys/rseq.h >> new file mode 100644 >> index 0000000000..503dce4cac >> --- /dev/null >> +++ b/sysdeps/unix/sysv/linux/sys/rseq.h >> @@ -0,0 +1,186 @@ > > I think there is some value in making this header compatible with > inclusion from the assembler (including constants for the relevant > struct offsets), but that can be a later change. Agreed. By "later", do you mean before merging the patch, between merge of the patch and next glibc release, or for a subsequent glibc release ? > >> +#ifndef _SYS_RSEQ_H >> +#define _SYS_RSEQ_H 1 >> + >> +/* Architecture-specific rseq signature. */ >> +#include <bits/rseq.h> > > Maybe add a newline between the above and the following, to make clear > the comment only applies to the first #include. OK > >> +#include <stdint.h> >> +#include <sys/cdefs.h> >> + >> +#ifdef __has_include >> +# if __has_include ("linux/rseq.h") >> +# define __GLIBC_HAVE_KERNEL_RSEQ >> +# endif >> +#else >> +# include <linux/version.h> >> +# if LINUX_VERSION_CODE >= KERNEL_VERSION (4, 18, 0) >> +# define __GLIBC_HAVE_KERNEL_RSEQ >> +# endif >> +#endif >> + >> +#ifdef __GLIBC_HAVE_KERNEL_RSEQ >> +/* We use the structures declarations from the kernel headers. */ >> +# include <linux/rseq.h> >> +#else >> +/* We use a copy of the include/uapi/linux/rseq.h kernel header. */ >> + >> +#include <asm/byteorder.h> >> + >> +enum rseq_cpu_id_state >> + { >> + RSEQ_CPU_ID_UNINITIALIZED = -1, >> + RSEQ_CPU_ID_REGISTRATION_FAILED = -2, >> + }; >> + >> +enum rseq_flags >> + { >> + RSEQ_FLAG_UNREGISTER = (1 << 0), >> + }; >> + >> +enum rseq_cs_flags_bit >> + { >> + RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT = 0, >> + RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT = 1, >> + RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT = 2, >> + }; >> + >> +enum rseq_cs_flags >> + { >> + RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT = >> + (1U << RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT), >> + RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL = >> + (1U << RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT), >> + RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE = >> + (1U << RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT), >> + }; >> + >> +/* struct rseq_cs is aligned on 4 * 8 bytes to ensure it is always >> + contained within a single cache-line. It is usually declared as >> + link-time constant data. */ >> +struct rseq_cs >> + { >> + /* Version of this structure. */ >> + uint32_t version; >> + /* enum rseq_cs_flags. */ >> + uint32_t flags; >> + uint64_t start_ip; >> + /* Offset from start_ip. */ >> + uint64_t post_commit_offset; >> + uint64_t abort_ip; >> +} __attribute__((aligned(4 * sizeof(uint64_t)))); > > The comment is wrong. 32-byte alignment does not put struct rseq_cs > on its own cache line on many (most?) CPUs. Not using the constant 32 > looks like unnecessary obfuscation to me. There is a difference between "being contained within a single cache-line" and "being the only structure in a cache-line". The objective here is the former. For instance, if we do not enforce any minimum alignment, the compiler could decide to align that structure on __alignof__(uint64_t) which happens to be 4 bytes on some architectures. This can cause this frequently accessed structure to be sitting across 2 cache-lines, thus requiring the CPU to load two cache lines rather than one very frequently. I think what you have in mind is "being the only structure in a cache-line", which is useful to eliminate false-sharing. However, considering that this is a TLS variable, we don't care about false-sharing, because it is never meant to be updated concurrently by many threads or CPUs. So I think my comment is correct. I agree that the constant 32 may be clearer here. I will change to align(32). > I still think we should avoid the alignment. The _ip fields should > perhaps be _pc (IP is more or less specific to x86). I am concerned that removing an alignment attribute which is exposed in a public Linux UAPI header can be an ABI breakage. I am also concerned about changing field names for fields already exposed in a public Linux UAPI header, especially if the change is only for cosmetic reasons. > > { and } are not aligned. Please do not forget to add spaces before > opening parentheses, and two spaces after the . and the end of > sentences. The opening { should always be on its own line. (This > also applies to the definition of struct rseq below.) OK. Old coding style habits die hard ;) > >> + >> +/* struct rseq is aligned on 4 * 8 bytes to ensure it is always >> + contained within a single cache-line. >> + >> + A single struct rseq per thread is allowed. */ >> +struct rseq >> + { >> + /* Restartable sequences cpu_id_start field. Updated by the >> + kernel. Read by user-space with single-copy atomicity >> + semantics. This field should only be read by the thread which >> + registered this data structure. Aligned on 32-bit. Always > > What does “Aligned on 32-bit” mean in this context? Do you mean to > reference 32-*byte* alignment here? No. I really mean 32-bit (4-byte). Being aligned on 32-byte guarantees that this field is aligned at least on 4-byte. This is required by single-copy atomicity semantics. Should I update this comment to state "Aligned on 4-byte" instead ? > >> + /* Restartable sequences rseq_cs field. >> + >> + Contains NULL when no critical section is active for the current >> + thread, or holds a pointer to the currently active struct rseq_cs. >> + >> + Updated by user-space, which sets the address of the currently >> + active rseq_cs at the beginning of assembly instruction sequence >> + block, and set to NULL by the kernel when it restarts an assembly >> + instruction sequence block, as well as when the kernel detects that >> + it is preempting or delivering a signal outside of the range >> + targeted by the rseq_cs. Also needs to be set to NULL by user-space >> + before reclaiming memory that contains the targeted struct rseq_cs. >> + >> + Read and set by the kernel. Set by user-space with single-copy >> + atomicity semantics. This field should only be updated by the >> + thread which registered this data structure. Aligned on 64-bit. */ >> + union { >> + uint64_t ptr64; >> +#ifdef __LP64__ >> + uint64_t ptr; >> +#else >> + struct { >> +#if (defined(__BYTE_ORDER) && (__BYTE_ORDER == __BIG_ENDIAN)) || >> defined(__BIG_ENDIAN) >> + uint32_t padding; /* Initialized to zero. */ >> + uint32_t ptr32; >> +#else /* LITTLE */ >> + uint32_t ptr32; >> + uint32_t padding; /* Initialized to zero. */ >> +#endif /* ENDIAN */ >> + } ptr; >> +#endif >> + } rseq_cs; > > Are these conditionals correct for x32? Let's see. With x86 gcc: -m64: (__x86_64__ && __LP64__) -m32: (__i386__) -mx32: (__x86_64__ && __ILP32__) So with "#ifdef __LP64__" we specifically target 64-bit pointers. The rest falls into the "else" case, which expects 32-bit pointers. Considering that x32 has 32-bit pointers, I don't see any issue here. > Shouldn't there be a member of type const struct rseq_cs * somewhere? Having pointers within structures in kernel UAPI headers is frowned upon. Indeed here having it in the union could possibly make some use-cases easier in user-space, so I'm open to it. It basically depends on how much we want the Linux UAPI header and the glibc header to stay in sync, and if other kernel maintainers are open to this addition. We don't mind that user-space uses that pointer, but we never want the kernel to touch that pointer rather than the 32/64-bit-aware fields. One possibility would be to do: union { uint64_t ptr64; #ifdef __LP64__ uint64_t ptr; #else struct { #if (defined (__BYTE_ORDER) && (__BYTE_ORDER == __BIG_ENDIAN)) || defined (__BIG_ENDIAN) uint32_t padding; /* Initialized to zero. */ uint32_t ptr32; #else /* LITTLE */ uint32_t ptr32; uint32_t padding; /* Initialized to zero. */ #endif /* ENDIAN */ } ptr; #endif #ifndef __KERNEL__ const struct rseq_cs *uptr; #endif } rseq_cs; in the union, so only user-space can see that field. Thoughts ? > >> diff --git a/sysdeps/unix/sysv/linux/x86/bits/rseq.h >> b/sysdeps/unix/sysv/linux/x86/bits/rseq.h >> new file mode 100644 >> index 0000000000..75f52d9788 >> --- /dev/null >> +++ b/sysdeps/unix/sysv/linux/x86/bits/rseq.h >> @@ -0,0 +1,30 @@ >> +/* Restartable Sequences Linux x86 architecture header. >> + Copyright (C) 2019-2020 Free Software Foundation, Inc. > > Please make sure that none of the new files reference the 2019 year. > It should be 2020, per GNU policy. OK > > The patch needs some rebasing on top of current master. Done. Thanks for the review! Mathieu
----- On Apr 27, 2020, at 7:59 AM, Florian Weimer fw@deneb.enyo.de wrote: > * Mathieu Desnoyers via Libc-alpha: > >> diff --git a/elf/libc_early_init.c b/elf/libc_early_init.c >> index 1ac66d895d..30466afea0 100644 >> --- a/elf/libc_early_init.c >> +++ b/elf/libc_early_init.c >> @@ -18,10 +18,13 @@ >> >> #include <ctype.h> >> #include <libc-early-init.h> >> +#include <rseq-internal.h> >> >> void >> __libc_early_init (void) >> { >> /* Initialize ctype data. */ >> __ctype_init (); >> + /* Register rseq ABI to the kernel. */ >> + (void) rseq_register_current_thread (); >> } > > I think the registration must be restricted to the primary namespace. > Otherwise, LD_AUDIT will register the area to the secondary libc (in > the audit module), not the primary libc for the entire process. > > I think the easiest way to implement this for now is a flag argument > for __libc_early_init (as the upstream __libc_multiple_libcs is not > entirely accurate). I will submit a patch. OK, once I get the patch, I will pick it up in my series. Thanks, Mathieu
* Mathieu Desnoyers: >>> +#include <sys/syscall.h> >>> +#include <stdint.h> >>> +#include <kernel-features.h> >>> +#include <sys/rseq.h> >>> + >>> +__thread struct rseq __rseq_abi = { >>> + .cpu_id = RSEQ_CPU_ID_UNINITIALIZED, >>> +}; >> >> { should go onto its own line. > > OK > >> I'd also add attribute_tls_model_ie, >> also it's implied by the declaration in the header. > > This contradicts feedback I received from Szabolcs Nagy in September 2019: > > https://public-inbox.org/libc-alpha/c58d4d6e-f22a-f5d9-e23a-5bd72cec1a86@arm.com/ > > "note that libpthread.so is built with -ftls-model=initial-exec > > (and if it wasn't then you'd want to put the attribute on the > declaration in the internal header file, not on the definition, > so the actual tls accesses generate the right code)" > > In the context of his feedback, __rseq_abi was defined within nptl/pthread_create.c. > It is now defined in sysdeps/unix/sysv/linux/rseq-sym.c, which is built into the > csu which AFAIU ends up in libc.so. His comment still applies though, because > libc.so is also built with -ftls-model=initial-exec. > > So should I apply the "initial-exec" TLS model only to the __rseq_abi > declaration, or is it preferred to apply it to both the declaration > and the definition ? I do not have a strong preference here. Technically, the declaration in the header file should be enough. >>> diff --git a/sysdeps/unix/sysv/linux/sys/rseq.h >>> b/sysdeps/unix/sysv/linux/sys/rseq.h >>> new file mode 100644 >>> index 0000000000..503dce4cac >>> --- /dev/null >>> +++ b/sysdeps/unix/sysv/linux/sys/rseq.h >>> @@ -0,0 +1,186 @@ >> >> I think there is some value in making this header compatible with >> inclusion from the assembler (including constants for the relevant >> struct offsets), but that can be a later change. > > Agreed. By "later", do you mean before merging the patch, between > merge of the patch and next glibc release, or for a subsequent glibc > release ? It can happen some time after merging the patch, preferably for this release. But I don't think it's release-critical. >>> +/* struct rseq_cs is aligned on 4 * 8 bytes to ensure it is always >>> + contained within a single cache-line. It is usually declared as >>> + link-time constant data. */ >>> +struct rseq_cs >>> + { >>> + /* Version of this structure. */ >>> + uint32_t version; >>> + /* enum rseq_cs_flags. */ >>> + uint32_t flags; >>> + uint64_t start_ip; >>> + /* Offset from start_ip. */ >>> + uint64_t post_commit_offset; >>> + uint64_t abort_ip; >>> +} __attribute__((aligned(4 * sizeof(uint64_t)))); >> >> The comment is wrong. 32-byte alignment does not put struct rseq_cs >> on its own cache line on many (most?) CPUs. Not using the constant 32 >> looks like unnecessary obfuscation to me. > > There is a difference between "being contained within a single cache-line" > and "being the only structure in a cache-line". The objective here is the > former. Fair enough. > I agree that the constant 32 may be clearer here. I will change to align(32). With a space, please. 8-) >>> +/* struct rseq is aligned on 4 * 8 bytes to ensure it is always >>> + contained within a single cache-line. >>> + >>> + A single struct rseq per thread is allowed. */ >>> +struct rseq >>> + { >>> + /* Restartable sequences cpu_id_start field. Updated by the >>> + kernel. Read by user-space with single-copy atomicity >>> + semantics. This field should only be read by the thread which >>> + registered this data structure. Aligned on 32-bit. Always >> >> What does “Aligned on 32-bit” mean in this context? Do you mean to >> reference 32-*byte* alignment here? > > No. I really mean 32-bit (4-byte). Being aligned on 32-byte guarantees that > this field is aligned at least on 4-byte. This is required by single-copy > atomicity semantics. > > Should I update this comment to state "Aligned on 4-byte" instead ? I think this is implied by all Linux ABIs. And the explicit alignment specification for struct rseq makes the alignment 32 bytes. >>> + /* Restartable sequences rseq_cs field. >>> + >>> + Contains NULL when no critical section is active for the current >>> + thread, or holds a pointer to the currently active struct rseq_cs. >>> + >>> + Updated by user-space, which sets the address of the currently >>> + active rseq_cs at the beginning of assembly instruction sequence >>> + block, and set to NULL by the kernel when it restarts an assembly >>> + instruction sequence block, as well as when the kernel detects that >>> + it is preempting or delivering a signal outside of the range >>> + targeted by the rseq_cs. Also needs to be set to NULL by user-space >>> + before reclaiming memory that contains the targeted struct rseq_cs. >>> + >>> + Read and set by the kernel. Set by user-space with single-copy >>> + atomicity semantics. This field should only be updated by the >>> + thread which registered this data structure. Aligned on 64-bit. */ >>> + union { >>> + uint64_t ptr64; >>> +#ifdef __LP64__ >>> + uint64_t ptr; >>> +#else >>> + struct { >>> +#if (defined(__BYTE_ORDER) && (__BYTE_ORDER == __BIG_ENDIAN)) || >>> defined(__BIG_ENDIAN) >>> + uint32_t padding; /* Initialized to zero. */ >>> + uint32_t ptr32; >>> +#else /* LITTLE */ >>> + uint32_t ptr32; >>> + uint32_t padding; /* Initialized to zero. */ >>> +#endif /* ENDIAN */ >>> + } ptr; >>> +#endif >>> + } rseq_cs; >> >> Are these conditionals correct for x32? > > Let's see. With x86 gcc: > > -m64: (__x86_64__ && __LP64__) > -m32: (__i386__) > -mx32: (__x86_64__ && __ILP32__) > > So with "#ifdef __LP64__" we specifically target 64-bit pointers. The rest > falls into the "else" case, which expects 32-bit pointers. Considering that > x32 has 32-bit pointers, I don't see any issue here. Does the kernel have a separate 32-bit entry point for rseq on x32? If not, it will expect the 64-bit struct layout. > We don't mind that user-space uses that pointer, but we never want the kernel > to touch that pointer rather than the 32/64-bit-aware fields. One possibility > would be to do: > > union > { > uint64_t ptr64; > #ifdef __LP64__ > uint64_t ptr; > #else > struct > { > #if (defined (__BYTE_ORDER) && (__BYTE_ORDER == __BIG_ENDIAN)) || defined (__BIG_ENDIAN) > uint32_t padding; /* Initialized to zero. */ > uint32_t ptr32; > #else /* LITTLE */ > uint32_t ptr32; > uint32_t padding; /* Initialized to zero. */ > #endif /* ENDIAN */ > } ptr; > #endif > > #ifndef __KERNEL__ > const struct rseq_cs *uptr; > #endif > } rseq_cs; > > in the union, so only user-space can see that field. Thoughts ? I think this depends on where the x32 question lands.
* Mathieu Desnoyers: > ----- On Apr 27, 2020, at 7:59 AM, Florian Weimer fw@deneb.enyo.de wrote: > >> * Mathieu Desnoyers via Libc-alpha: >> >>> diff --git a/elf/libc_early_init.c b/elf/libc_early_init.c >>> index 1ac66d895d..30466afea0 100644 >>> --- a/elf/libc_early_init.c >>> +++ b/elf/libc_early_init.c >>> @@ -18,10 +18,13 @@ >>> >>> #include <ctype.h> >>> #include <libc-early-init.h> >>> +#include <rseq-internal.h> >>> >>> void >>> __libc_early_init (void) >>> { >>> /* Initialize ctype data. */ >>> __ctype_init (); >>> + /* Register rseq ABI to the kernel. */ >>> + (void) rseq_register_current_thread (); >>> } >> >> I think the registration must be restricted to the primary namespace. >> Otherwise, LD_AUDIT will register the area to the secondary libc (in >> the audit module), not the primary libc for the entire process. >> >> I think the easiest way to implement this for now is a flag argument >> for __libc_early_init (as the upstream __libc_multiple_libcs is not >> entirely accurate). I will submit a patch. > > OK, once I get the patch, I will pick it up in my series. There should be no need for that, it can be reviewed and committed separately: <https://sourceware.org/pipermail/libc-alpha/2020-April/113182.html>
----- On Apr 27, 2020, at 12:54 PM, Florian Weimer fw@deneb.enyo.de wrote: > * Mathieu Desnoyers: > >>>> +#include <sys/syscall.h> >>>> +#include <stdint.h> >>>> +#include <kernel-features.h> >>>> +#include <sys/rseq.h> >>>> + >>>> +__thread struct rseq __rseq_abi = { >>>> + .cpu_id = RSEQ_CPU_ID_UNINITIALIZED, >>>> +}; >>> >>> { should go onto its own line. >> >> OK >> >>> I'd also add attribute_tls_model_ie, >>> also it's implied by the declaration in the header. >> >> This contradicts feedback I received from Szabolcs Nagy in September 2019: >> >> https://public-inbox.org/libc-alpha/c58d4d6e-f22a-f5d9-e23a-5bd72cec1a86@arm.com/ >> >> "note that libpthread.so is built with -ftls-model=initial-exec >> >> (and if it wasn't then you'd want to put the attribute on the >> declaration in the internal header file, not on the definition, >> so the actual tls accesses generate the right code)" >> >> In the context of his feedback, __rseq_abi was defined within >> nptl/pthread_create.c. >> It is now defined in sysdeps/unix/sysv/linux/rseq-sym.c, which is built into the >> csu which AFAIU ends up in libc.so. His comment still applies though, because >> libc.so is also built with -ftls-model=initial-exec. >> >> So should I apply the "initial-exec" TLS model only to the __rseq_abi >> declaration, or is it preferred to apply it to both the declaration >> and the definition ? > > I do not have a strong preference here. Technically, the declaration > in the header file should be enough. OK, so I'll just keep the attribute on the declaration in the header. > >>>> diff --git a/sysdeps/unix/sysv/linux/sys/rseq.h >>>> b/sysdeps/unix/sysv/linux/sys/rseq.h >>>> new file mode 100644 >>>> index 0000000000..503dce4cac >>>> --- /dev/null >>>> +++ b/sysdeps/unix/sysv/linux/sys/rseq.h >>>> @@ -0,0 +1,186 @@ >>> >>> I think there is some value in making this header compatible with >>> inclusion from the assembler (including constants for the relevant >>> struct offsets), but that can be a later change. >> >> Agreed. By "later", do you mean before merging the patch, between >> merge of the patch and next glibc release, or for a subsequent glibc >> release ? > > It can happen some time after merging the patch, preferably for this > release. But I don't think it's release-critical. OK > >>>> +/* struct rseq is aligned on 4 * 8 bytes to ensure it is always >>>> + contained within a single cache-line. >>>> + >>>> + A single struct rseq per thread is allowed. */ >>>> +struct rseq >>>> + { >>>> + /* Restartable sequences cpu_id_start field. Updated by the >>>> + kernel. Read by user-space with single-copy atomicity >>>> + semantics. This field should only be read by the thread which >>>> + registered this data structure. Aligned on 32-bit. Always >>> >>> What does “Aligned on 32-bit” mean in this context? Do you mean to >>> reference 32-*byte* alignment here? >> >> No. I really mean 32-bit (4-byte). Being aligned on 32-byte guarantees that >> this field is aligned at least on 4-byte. This is required by single-copy >> atomicity semantics. >> >> Should I update this comment to state "Aligned on 4-byte" instead ? > > I think this is implied by all Linux ABIs. And the explicit alignment > specification for struct rseq makes the alignment 32 bytes. Unless a structure ends up being packed, which is of course not the case here. I would prefer to keep the comment about 32-bit alignment requirement on the specific fields, because the motivation for alignment requirement is much more strict for fields (correctness) than the motivation for alignment of the structure (performance). > >>>> + /* Restartable sequences rseq_cs field. >>>> + >>>> + Contains NULL when no critical section is active for the current >>>> + thread, or holds a pointer to the currently active struct rseq_cs. >>>> + >>>> + Updated by user-space, which sets the address of the currently >>>> + active rseq_cs at the beginning of assembly instruction sequence >>>> + block, and set to NULL by the kernel when it restarts an assembly >>>> + instruction sequence block, as well as when the kernel detects that >>>> + it is preempting or delivering a signal outside of the range >>>> + targeted by the rseq_cs. Also needs to be set to NULL by user-space >>>> + before reclaiming memory that contains the targeted struct rseq_cs. >>>> + >>>> + Read and set by the kernel. Set by user-space with single-copy >>>> + atomicity semantics. This field should only be updated by the >>>> + thread which registered this data structure. Aligned on 64-bit. */ >>>> + union { >>>> + uint64_t ptr64; >>>> +#ifdef __LP64__ >>>> + uint64_t ptr; >>>> +#else >>>> + struct { >>>> +#if (defined(__BYTE_ORDER) && (__BYTE_ORDER == __BIG_ENDIAN)) || >>>> defined(__BIG_ENDIAN) >>>> + uint32_t padding; /* Initialized to zero. */ >>>> + uint32_t ptr32; >>>> +#else /* LITTLE */ >>>> + uint32_t ptr32; >>>> + uint32_t padding; /* Initialized to zero. */ >>>> +#endif /* ENDIAN */ >>>> + } ptr; >>>> +#endif >>>> + } rseq_cs; >>> >>> Are these conditionals correct for x32? >> >> Let's see. With x86 gcc: >> >> -m64: (__x86_64__ && __LP64__) >> -m32: (__i386__) >> -mx32: (__x86_64__ && __ILP32__) >> >> So with "#ifdef __LP64__" we specifically target 64-bit pointers. The rest >> falls into the "else" case, which expects 32-bit pointers. Considering that >> x32 has 32-bit pointers, I don't see any issue here. > > Does the kernel have a separate 32-bit entry point for rseq on x32? > If not, it will expect the 64-bit struct layout. No, there is a single entry point into rseq covering all of 32-bit, 64-bit and x32. We achieve this by ensuring the layout of the linux/rseq.h structures uses the union representation for pointers. Therefore, the kernel does not care whether it reads a pointer from a 32-bit or 64-bit process. This is becoming the preferred way to design Linux kernel ABIs nowadays. > >> We don't mind that user-space uses that pointer, but we never want the kernel >> to touch that pointer rather than the 32/64-bit-aware fields. One possibility >> would be to do: >> >> union >> { >> uint64_t ptr64; >> #ifdef __LP64__ >> uint64_t ptr; >> #else >> struct >> { >> #if (defined (__BYTE_ORDER) && (__BYTE_ORDER == __BIG_ENDIAN)) || defined >> (__BIG_ENDIAN) >> uint32_t padding; /* Initialized to zero. */ >> uint32_t ptr32; >> #else /* LITTLE */ >> uint32_t ptr32; >> uint32_t padding; /* Initialized to zero. */ >> #endif /* ENDIAN */ >> } ptr; >> #endif >> >> #ifndef __KERNEL__ >> const struct rseq_cs *uptr; >> #endif >> } rseq_cs; >> >> in the union, so only user-space can see that field. Thoughts ? > > I think this depends on where the x32 question lands. x32 should not be an issue as explained above, so I'm very open to add this "uptr" for user-space only. Thanks, Mathieu
----- On Apr 27, 2020, at 1:26 PM, Mathieu Desnoyers mathieu.desnoyers@efficios.com wrote: [...] >>> We don't mind that user-space uses that pointer, but we never want the kernel >>> to touch that pointer rather than the 32/64-bit-aware fields. One possibility >>> would be to do: >>> >>> union >>> { >>> uint64_t ptr64; >>> #ifdef __LP64__ >>> uint64_t ptr; >>> #else >>> struct >>> { >>> #if (defined (__BYTE_ORDER) && (__BYTE_ORDER == __BIG_ENDIAN)) || defined >>> (__BIG_ENDIAN) >>> uint32_t padding; /* Initialized to zero. */ >>> uint32_t ptr32; >>> #else /* LITTLE */ >>> uint32_t ptr32; >>> uint32_t padding; /* Initialized to zero. */ >>> #endif /* ENDIAN */ >>> } ptr; >>> #endif >>> >>> #ifndef __KERNEL__ >>> const struct rseq_cs *uptr; >>> #endif >>> } rseq_cs; >>> >>> in the union, so only user-space can see that field. Thoughts ? >> >> I think this depends on where the x32 question lands. > > x32 should not be an issue as explained above, so I'm very open to > add this "uptr" for user-space only. Actually, the snippet above is broken on 32-bit. It needs to be: union { uint64_t ptr64; #ifdef __LP64__ uint64_t ptr; # ifndef __KERNEL__ const struct rseq_cs *uptr; # endif #else struct { #if (defined (__BYTE_ORDER) && (__BYTE_ORDER == __BIG_ENDIAN)) || defined (__BIG_ENDIAN) uint32_t padding; /* Initialized to zero. */ uint32_t ptr32; #else /* LITTLE */ uint32_t ptr32; uint32_t padding; /* Initialized to zero. */ #endif /* ENDIAN */ } ptr; # ifndef __KERNEL__ struct { # if (defined (__BYTE_ORDER) && (__BYTE_ORDER == __BIG_ENDIAN)) || defined (__BIG_ENDIAN) uint32_t padding; /* Initialized to zero. */ const struct rseq_cs *uptr32; # else /* LITTLE */ const struct rseq_cs *uptr32; uint32_t padding; /* Initialized to zero. */ # endif /* ENDIAN */ } uptr; # endif #endif } rseq_cs; I'll leave this out of the patchset for now as we'd need more feedback on its usefulness. Thanks, Mathieu
* Mathieu Desnoyers: >>>>> +/* struct rseq is aligned on 4 * 8 bytes to ensure it is always >>>>> + contained within a single cache-line. >>>>> + >>>>> + A single struct rseq per thread is allowed. */ >>>>> +struct rseq >>>>> + { >>>>> + /* Restartable sequences cpu_id_start field. Updated by the >>>>> + kernel. Read by user-space with single-copy atomicity >>>>> + semantics. This field should only be read by the thread which >>>>> + registered this data structure. Aligned on 32-bit. Always >>>> >>>> What does “Aligned on 32-bit” mean in this context? Do you mean to >>>> reference 32-*byte* alignment here? >>> >>> No. I really mean 32-bit (4-byte). Being aligned on 32-byte guarantees that >>> this field is aligned at least on 4-byte. This is required by single-copy >>> atomicity semantics. >>> >>> Should I update this comment to state "Aligned on 4-byte" instead ? >> >> I think this is implied by all Linux ABIs. And the explicit alignment >> specification for struct rseq makes the alignment 32 bytes. > > Unless a structure ends up being packed, which is of course not the case > here. > > I would prefer to keep the comment about 32-bit alignment requirement on > the specific fields, because the motivation for alignment requirement is > much more strict for fields (correctness) than the motivation for alignment > of the structure (performance). But the correctness is already enforced by the compiler, so I fail to see point of mentioning this in the comment. Anyway, I don't want to make a big deal of it. Please leave it in if you think it is ehlpful. > x32 should not be an issue as explained above, so I'm very open to > add this "uptr" for user-space only. Okay, then please use anonymous unions and structs as necessary, to ensure that the uptr field can be reached on all platforms in the same way.
----- On Apr 28, 2020, at 8:02 AM, Florian Weimer fw@deneb.enyo.de wrote: > * Mathieu Desnoyers: > >>>>>> +/* struct rseq is aligned on 4 * 8 bytes to ensure it is always >>>>>> + contained within a single cache-line. >>>>>> + >>>>>> + A single struct rseq per thread is allowed. */ >>>>>> +struct rseq >>>>>> + { >>>>>> + /* Restartable sequences cpu_id_start field. Updated by the >>>>>> + kernel. Read by user-space with single-copy atomicity >>>>>> + semantics. This field should only be read by the thread which >>>>>> + registered this data structure. Aligned on 32-bit. Always >>>>> >>>>> What does “Aligned on 32-bit” mean in this context? Do you mean to >>>>> reference 32-*byte* alignment here? >>>> >>>> No. I really mean 32-bit (4-byte). Being aligned on 32-byte guarantees that >>>> this field is aligned at least on 4-byte. This is required by single-copy >>>> atomicity semantics. >>>> >>>> Should I update this comment to state "Aligned on 4-byte" instead ? >>> >>> I think this is implied by all Linux ABIs. And the explicit alignment >>> specification for struct rseq makes the alignment 32 bytes. >> >> Unless a structure ends up being packed, which is of course not the case >> here. >> >> I would prefer to keep the comment about 32-bit alignment requirement on >> the specific fields, because the motivation for alignment requirement is >> much more strict for fields (correctness) than the motivation for alignment >> of the structure (performance). > > But the correctness is already enforced by the compiler, so I fail to > see point of mentioning this in the comment. > > Anyway, I don't want to make a big deal of it. Please leave it in if > you think it is ehlpful. I would prefer to leave it in, just to make the requirements plain clear in case those structures are allocated on the heap (for instance). > >> x32 should not be an issue as explained above, so I'm very open to >> add this "uptr" for user-space only. > > Okay, then please use anonymous unions and structs as necessary, to > ensure that the uptr field can be reached on all platforms in the same > way. OK, will do! One issue I'm currently facing when running "make check": because nptl/tst-rseq-nptl.c uses pthread_cancel(), I run into an Abort with: libgcc_s.so.1 must be installed for pthread_cancel to work Didn't expect signal from child: got `Aborted' So far I've tested the rest of that file with a patch on top which disables the use of pthread_cancel (), but I'd really like to give it a full coverage before sending this out. In https://sourceware.org/glibc/wiki/Testing/Builds there is a section about "Building glibc with intent to install" which describes that libgcc must be copied manually. My use-case is that I just want to run "make check" in the build directory and make sure it finds the libgcc it needs to succeed using pthread_cancel (). How can I achieve this ? Thanks, Mathieu
* Mathieu Desnoyers: > One issue I'm currently facing when running "make check": because > nptl/tst-rseq-nptl.c uses pthread_cancel(), I run into an Abort > with: > > libgcc_s.so.1 must be installed for pthread_cancel to work > Didn't expect signal from child: got `Aborted' This is really unusual. Is the affected test statically linked?
----- On Apr 28, 2020, at 8:35 AM, Florian Weimer fw@deneb.enyo.de wrote: > * Mathieu Desnoyers: > >> One issue I'm currently facing when running "make check": because >> nptl/tst-rseq-nptl.c uses pthread_cancel(), I run into an Abort >> with: >> >> libgcc_s.so.1 must be installed for pthread_cancel to work >> Didn't expect signal from child: got `Aborted' > > This is really unusual. Is the affected test statically linked? I built glibc without specifying anything particular, and ran "make check". It indeed seems to be dynamically linked to libc: ldd tst-rseq-nptl ./tst-rseq-nptl: /lib/x86_64-linux-gnu/libc.so.6: version `GLIBC_2.32' not found (required by ./tst-rseq-nptl) linux-vdso.so.1 (0x00007ffd3a2f3000) libpthread.so.0 => /lib/x86_64-linux-gnu/libpthread.so.0 (0x00007f0527560000) libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007f052716f000) /home/efficios/glibc-test5/lib/ld-linux-x86-64.so.2 => /lib64/ld-linux-x86-64.so.2 (0x00007f0527986000) After make check I have: cat tst-rseq-nptl.test-result FAIL: nptl/tst-rseq-nptl original exit status 134 And if I run ./tst-rseq-nptl Then I get libgcc_s.so.1 must be installed for pthread_cancel to work Didn't expect signal from child: got `Aborted' libgcc_s.so.1 must be installed for pthread_cancel to work Aborted (core dumped) Same result if I do ./testrun.sh nptl/tst-rseq-nptl Thanks, Mathieu
* Mathieu Desnoyers: > ----- On Apr 28, 2020, at 8:35 AM, Florian Weimer fw@deneb.enyo.de wrote: > >> * Mathieu Desnoyers: >> >>> One issue I'm currently facing when running "make check": because >>> nptl/tst-rseq-nptl.c uses pthread_cancel(), I run into an Abort >>> with: >>> >>> libgcc_s.so.1 must be installed for pthread_cancel to work >>> Didn't expect signal from child: got `Aborted' >> >> This is really unusual. Is the affected test statically linked? > > I built glibc without specifying anything particular, and ran > "make check". It indeed seems to be dynamically linked to libc: > > ldd tst-rseq-nptl > ./tst-rseq-nptl: /lib/x86_64-linux-gnu/libc.so.6: version `GLIBC_2.32' not found (required by ./tst-rseq-nptl) > linux-vdso.so.1 (0x00007ffd3a2f3000) > libpthread.so.0 => /lib/x86_64-linux-gnu/libpthread.so.0 (0x00007f0527560000) > libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007f052716f000) > /home/efficios/glibc-test5/lib/ld-linux-x86-64.so.2 => /lib64/ld-linux-x86-64.so.2 (0x00007f0527986000) That's expected if the installed glibc is older than the built glibc. > After make check I have: > > cat tst-rseq-nptl.test-result > FAIL: nptl/tst-rseq-nptl > original exit status 134 What's in the tst-rseq-nptl.out file? > And if I run > > ./tst-rseq-nptl > > Then I get > > libgcc_s.so.1 must be installed for pthread_cancel to work > Didn't expect signal from child: got `Aborted' > libgcc_s.so.1 must be installed for pthread_cancel to work > Aborted (core dumped) I'm puzzled why you don't get a GLIBC_2.32 version error in this case. Do you build with --enable-hardcoded-path-in-tests? > Same result if I do ./testrun.sh nptl/tst-rseq-nptl That one definitely should work. I expect you might see this if libgcc_s.so.1 is installed into a multiarch subdirectory that upstream glibc does not search. (The Debian patches are unfortunately not upstream.) I think on my system, the built glibc can find the system libgcc_s via /etc/ld.so.cache, so I haven't seen this issue yet.
----- On Apr 28, 2020, at 8:33 AM, Mathieu Desnoyers mathieu.desnoyers@efficios.com wrote: > ----- On Apr 28, 2020, at 8:02 AM, Florian Weimer fw@deneb.enyo.de wrote: > [...] >> >>> x32 should not be an issue as explained above, so I'm very open to >>> add this "uptr" for user-space only. >> >> Okay, then please use anonymous unions and structs as necessary, to >> ensure that the uptr field can be reached on all platforms in the same >> way. > > OK, will do! What I came up with looks like this. User-space can use rseq_cs.uptr.ptr both on 32-bit and 64-bit to update the pointer: /* Restartable sequences rseq_cs field. Contains NULL when no critical section is active for the current thread, or holds a pointer to the currently active struct rseq_cs. Updated by user-space, which sets the address of the currently active rseq_cs at the beginning of assembly instruction sequence block, and set to NULL by the kernel when it restarts an assembly instruction sequence block, as well as when the kernel detects that it is preempting or delivering a signal outside of the range targeted by the rseq_cs. Also needs to be set to NULL by user-space before reclaiming memory that contains the targeted struct rseq_cs. Read and set by the kernel. Set by user-space with single-copy atomicity semantics. This field should only be updated by the thread which registered this data structure. Aligned on 64-bit. User-space may perform the update through the rseq_cs.uptr.ptr field. The padding needs to be initialized to zero on 32-bit. */ union { uint64_t ptr64; #ifdef __LP64__ uint64_t ptr; #else struct { # if (defined (__BYTE_ORDER) && (__BYTE_ORDER == __BIG_ENDIAN)) || defined (__BIG_ENDIAN) uint32_t padding; /* Initialized to zero. */ uint32_t ptr32; # else /* LITTLE */ uint32_t ptr32; uint32_t padding; /* Initialized to zero. */ # endif /* ENDIAN */ } ptr; #endif #ifndef __KERNEL__ struct { # ifdef __LP64__ const struct rseq_cs *ptr; # else # if (defined (__BYTE_ORDER) && (__BYTE_ORDER == __BIG_ENDIAN)) || defined (__BIG_ENDIAN) uint32_t padding; /* Initialized to zero. */ const struct rseq_cs *ptr; # else /* LITTLE */ const struct rseq_cs *ptr; uint32_t padding; /* Initialized to zero. */ # endif /* ENDIAN */ # endif } uptr; #endif } rseq_cs; Thanks, Mathieu
----- On Apr 28, 2020, at 8:54 AM, Florian Weimer fw@deneb.enyo.de wrote: > * Mathieu Desnoyers: > >> ----- On Apr 28, 2020, at 8:35 AM, Florian Weimer fw@deneb.enyo.de wrote: >> >>> * Mathieu Desnoyers: >>> >>>> One issue I'm currently facing when running "make check": because >>>> nptl/tst-rseq-nptl.c uses pthread_cancel(), I run into an Abort >>>> with: >>>> >>>> libgcc_s.so.1 must be installed for pthread_cancel to work >>>> Didn't expect signal from child: got `Aborted' >>> >>> This is really unusual. Is the affected test statically linked? >> >> I built glibc without specifying anything particular, and ran >> "make check". It indeed seems to be dynamically linked to libc: >> >> ldd tst-rseq-nptl >> ./tst-rseq-nptl: /lib/x86_64-linux-gnu/libc.so.6: version `GLIBC_2.32' not found >> (required by ./tst-rseq-nptl) >> linux-vdso.so.1 (0x00007ffd3a2f3000) >> libpthread.so.0 => /lib/x86_64-linux-gnu/libpthread.so.0 (0x00007f0527560000) >> libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007f052716f000) >> /home/efficios/glibc-test5/lib/ld-linux-x86-64.so.2 => >> /lib64/ld-linux-x86-64.so.2 (0x00007f0527986000) > > That's expected if the installed glibc is older than the built glibc. > >> After make check I have: >> >> cat tst-rseq-nptl.test-result >> FAIL: nptl/tst-rseq-nptl >> original exit status 134 > > What's in the tst-rseq-nptl.out file? It contains: Didn't expect signal from child: got `Aborted' >> And if I run >> >> ./tst-rseq-nptl >> >> Then I get >> >> libgcc_s.so.1 must be installed for pthread_cancel to work >> Didn't expect signal from child: got `Aborted' >> libgcc_s.so.1 must be installed for pthread_cancel to work >> Aborted (core dumped) > > I'm puzzled why you don't get a GLIBC_2.32 version error in this case. > Do you build with --enable-hardcoded-path-in-tests? Just tried with this, and it fails the same way. However, the output of ldd nptl/tst-rseq-nptl changes: linux-vdso.so.1 (0x00007ffc235c9000) libpthread.so.0 => /home/efficios/git/glibc-build/nptl/libpthread.so.0 (0x00007fd308278000) libc.so.6 => /home/efficios/git/glibc-build/libc.so.6 (0x00007fd307ebc000) /home/efficios/git/glibc-build/elf/ld.so => /lib64/ld-linux-x86-64.so.2 (0x00007fd30869e000) >> Same result if I do ./testrun.sh nptl/tst-rseq-nptl > > That one definitely should work. > > I expect you might see this if libgcc_s.so.1 is installed into a > multiarch subdirectory that upstream glibc does not search. (The > Debian patches are unfortunately not upstream.) My test environment is a Ubuntu 18.04.1 LTS. > > I think on my system, the built glibc can find the system libgcc_s via > /etc/ld.so.cache, so I haven't seen this issue yet. On my system, libgcc_s is provided here: /lib/x86_64-linux-gnu/libgcc_s.so.1 by this package: Package: libgcc1 Architecture: amd64 Version: 1:8.4.0-1ubuntu1~18.04 Thanks, Mathieu
The 04/28/2020 10:58, Mathieu Desnoyers wrote: > ----- On Apr 28, 2020, at 8:54 AM, Florian Weimer fw@deneb.enyo.de wrote: > > That one definitely should work. > > > > I expect you might see this if libgcc_s.so.1 is installed into a > > multiarch subdirectory that upstream glibc does not search. (The > > Debian patches are unfortunately not upstream.) > > My test environment is a Ubuntu 18.04.1 LTS. > > > > > I think on my system, the built glibc can find the system libgcc_s via > > /etc/ld.so.cache, so I haven't seen this issue yet. > > On my system, libgcc_s is provided here: > > /lib/x86_64-linux-gnu/libgcc_s.so.1 > > by this package: > > Package: libgcc1 > Architecture: amd64 > Version: 1:8.4.0-1ubuntu1~18.04 before running the tests cp `$CC --print-file-name libgcc_s.so.1` glibc/build/dir cp `$CC --print-file-name libstdc++.so.6` glibc/build/dir so those toolchain libs are in the search path of the newly built libc when running tests.
* Szabolcs Nagy: > The 04/28/2020 10:58, Mathieu Desnoyers wrote: >> ----- On Apr 28, 2020, at 8:54 AM, Florian Weimer fw@deneb.enyo.de wrote: >> > That one definitely should work. >> > >> > I expect you might see this if libgcc_s.so.1 is installed into a >> > multiarch subdirectory that upstream glibc does not search. (The >> > Debian patches are unfortunately not upstream.) >> >> My test environment is a Ubuntu 18.04.1 LTS. >> >> > >> > I think on my system, the built glibc can find the system libgcc_s via >> > /etc/ld.so.cache, so I haven't seen this issue yet. >> >> On my system, libgcc_s is provided here: >> >> /lib/x86_64-linux-gnu/libgcc_s.so.1 >> >> by this package: >> >> Package: libgcc1 >> Architecture: amd64 >> Version: 1:8.4.0-1ubuntu1~18.04 > > before running the tests > > cp `$CC --print-file-name libgcc_s.so.1` glibc/build/dir > cp `$CC --print-file-name libstdc++.so.6` glibc/build/dir > > so those toolchain libs are in the search path > of the newly built libc when running tests. Do you actually see the need for these steps yourself? I guess the correct fix would be to upstream the Debian multiarch changes and activate them automatically with a configure check on systems that use multiarch paths.
The 04/29/2020 10:18, Florian Weimer wrote: > * Szabolcs Nagy: > > > The 04/28/2020 10:58, Mathieu Desnoyers wrote: > >> ----- On Apr 28, 2020, at 8:54 AM, Florian Weimer fw@deneb.enyo.de wrote: > >> > That one definitely should work. > >> > > >> > I expect you might see this if libgcc_s.so.1 is installed into a > >> > multiarch subdirectory that upstream glibc does not search. (The > >> > Debian patches are unfortunately not upstream.) > >> > >> My test environment is a Ubuntu 18.04.1 LTS. > >> > >> > > >> > I think on my system, the built glibc can find the system libgcc_s via > >> > /etc/ld.so.cache, so I haven't seen this issue yet. > >> > >> On my system, libgcc_s is provided here: > >> > >> /lib/x86_64-linux-gnu/libgcc_s.so.1 > >> > >> by this package: > >> > >> Package: libgcc1 > >> Architecture: amd64 > >> Version: 1:8.4.0-1ubuntu1~18.04 > > > > before running the tests > > > > cp `$CC --print-file-name libgcc_s.so.1` glibc/build/dir > > cp `$CC --print-file-name libstdc++.so.6` glibc/build/dir > > > > so those toolchain libs are in the search path > > of the newly built libc when running tests. > > Do you actually see the need for these steps yourself? > > I guess the correct fix would be to upstream the Debian multiarch > changes and activate them automatically with a configure check on > systems that use multiarch paths. cancel tests work for me on an ubuntu system because of /etc/ld.so.cache, but that may not be present or the system may not be glibc based at all. i always do the cp because i build gcc myself (usually close to current master) and don't install it to the system path which means at compile time and runtime different libraries are used if i dont copy
* Szabolcs Nagy: > cancel tests work for me on an ubuntu system because > of /etc/ld.so.cache, but that may not be present > or the system may not be glibc based at all. I see. > i always do the cp because i build gcc myself (usually > close to current master) and don't install it to the > system path which means at compile time and runtime > different libraries are used if i dont copy I wonder if we should do this automatically (maybe using a symbolic link), at least of GCC and ld can find these shared objects. Maybe it's possible to tell ld to provide output that makes it easy to locate the actual objects in the file system? (Trimmed the Cc: list.)
* Mathieu Desnoyers: > ----- On Apr 28, 2020, at 8:33 AM, Mathieu Desnoyers mathieu.desnoyers@efficios.com wrote: > >> ----- On Apr 28, 2020, at 8:02 AM, Florian Weimer fw@deneb.enyo.de wrote: >> > [...] >>> >>>> x32 should not be an issue as explained above, so I'm very open to >>>> add this "uptr" for user-space only. >>> >>> Okay, then please use anonymous unions and structs as necessary, to >>> ensure that the uptr field can be reached on all platforms in the same >>> way. >> >> OK, will do! > > What I came up with looks like this. User-space can use rseq_cs.uptr.ptr > both on 32-bit and 64-bit to update the pointer: Agreed, this should work.
----- On Apr 29, 2020, at 5:01 AM, Florian Weimer fw@deneb.enyo.de wrote: > * Szabolcs Nagy: > >> cancel tests work for me on an ubuntu system because >> of /etc/ld.so.cache, but that may not be present >> or the system may not be glibc based at all. > > I see. > >> i always do the cp because i build gcc myself (usually >> close to current master) and don't install it to the >> system path which means at compile time and runtime >> different libraries are used if i dont copy > > I wonder if we should do this automatically (maybe using a symbolic > link), at least of GCC and ld can find these shared objects. Maybe > it's possible to tell ld to provide output that makes it easy to > locate the actual objects in the file system? > > (Trimmed the Cc: list.) For the records, here is the output I get here: $ gcc --print-file-name libgcc_s.so.1 /usr/lib/gcc/x86_64-linux-gnu/7/libgcc_s.so.1 $ gcc --print-file-name libstdc++.so.6 /usr/lib/gcc/x86_64-linux-gnu/7/../../../x86_64-linux-gnu/libstdc++.so.6 $ gcc -v Using built-in specs. COLLECT_GCC=gcc COLLECT_LTO_WRAPPER=/usr/lib/gcc/x86_64-linux-gnu/7/lto-wrapper OFFLOAD_TARGET_NAMES=nvptx-none OFFLOAD_TARGET_DEFAULT=1 Target: x86_64-linux-gnu Configured with: ../src/configure -v --with-pkgversion='Ubuntu 7.5.0-3ubuntu1~18.04' --with-bugurl=file:///usr/share/doc/gcc-7/README.Bugs --enable-languages=c,ada,c++,go,brig,d,fortran,objc,obj-c++ --prefix=/usr --with-gcc-major-version-only --program-suffix=-7 --program-prefix=x86_64-linux-gnu- --enable-shared --enable-linker-build-id --libexecdir=/usr/lib --without-included-gettext --enable-threads=posix --libdir=/usr/lib --enable-nls --enable-bootstrap --enable-clocale=gnu --enable-libstdcxx-debug --enable-libstdcxx-time=yes --with-default-libstdcxx-abi=new --enable-gnu-unique-object --disable-vtable-verify --enable-libmpx --enable-plugin --enable-default-pie --with-system-zlib --with-target-system-zlib --enable-objc-gc=auto --enable-multiarch --disable-werror --with-arch-32=i686 --with-abi=m64 --with-multilib-list=m32,m64,mx32 --enable-multilib --with-tune=generic --enable-offload-targets=nvptx-none --without-cuda-driver --enable-checking=release --build=x86_64-linux-gnu --host=x86_64-linux-gnu --target=x86_64-linux-gnu Thread model: posix gcc version 7.5.0 (Ubuntu 7.5.0-3ubuntu1~18.04) Even if this becomes worked-around upstream, I suspect it would be good to mention this trick in https://sourceware.org/glibc/wiki/Testing/Builds for those working on older glibc versions. Considering that I my build directory is in: /home/efficios/git/glibc-build and I configure my glibc from the build directory with: "../glibc/configure --prefix=/home/efficios/glibc-test" Where exactly should I copy libgcc_s.so.1 and libstdc++.so.6 ? I tried copying them to /home/efficios/git/glibc-build/ but it does not appear to fix the issue. I notice that other .so are pulled into /home/efficios/git/glibc-build/testroot.pristine and then /home/efficios/git/glibc-build/testroot.root though. Is it where I should copy them, and if so, under which subdirectory are they expected ? Thanks, Mathieu
The 04/29/2020 08:57, Mathieu Desnoyers wrote: > ----- On Apr 29, 2020, at 5:01 AM, Florian Weimer fw@deneb.enyo.de wrote: > > > * Szabolcs Nagy: > > > >> cancel tests work for me on an ubuntu system because > >> of /etc/ld.so.cache, but that may not be present > >> or the system may not be glibc based at all. > > > > I see. > > > >> i always do the cp because i build gcc myself (usually > >> close to current master) and don't install it to the > >> system path which means at compile time and runtime > >> different libraries are used if i dont copy > > > > I wonder if we should do this automatically (maybe using a symbolic > > link), at least of GCC and ld can find these shared objects. Maybe > > it's possible to tell ld to provide output that makes it easy to > > locate the actual objects in the file system? > > > > (Trimmed the Cc: list.) > > For the records, here is the output I get here: > > $ gcc --print-file-name libgcc_s.so.1 > /usr/lib/gcc/x86_64-linux-gnu/7/libgcc_s.so.1 > $ gcc --print-file-name libstdc++.so.6 > /usr/lib/gcc/x86_64-linux-gnu/7/../../../x86_64-linux-gnu/libstdc++.so.6 > $ gcc -v > Using built-in specs. > COLLECT_GCC=gcc > COLLECT_LTO_WRAPPER=/usr/lib/gcc/x86_64-linux-gnu/7/lto-wrapper > OFFLOAD_TARGET_NAMES=nvptx-none > OFFLOAD_TARGET_DEFAULT=1 > Target: x86_64-linux-gnu > Configured with: ../src/configure -v --with-pkgversion='Ubuntu 7.5.0-3ubuntu1~18.04' --with-bugurl=file:///usr/share/doc/gcc-7/README.Bugs --enable-languages=c,ada,c++,go,brig,d,fortran,objc,obj-c++ --prefix=/usr --with-gcc-major-version-only --program-suffix=-7 --program-prefix=x86_64-linux-gnu- --enable-shared --enable-linker-build-id --libexecdir=/usr/lib --without-included-gettext --enable-threads=posix --libdir=/usr/lib --enable-nls --enable-bootstrap --enable-clocale=gnu --enable-libstdcxx-debug --enable-libstdcxx-time=yes --with-default-libstdcxx-abi=new --enable-gnu-unique-object --disable-vtable-verify --enable-libmpx --enable-plugin --enable-default-pie --with-system-zlib --with-target-system-zlib --enable-objc-gc=auto --enable-multiarch --disable-werror --with-arch-32=i686 --with-abi=m64 --with-multilib-list=m32,m64,mx32 --enable-multilib --with-tune=generic --enable-offload-targets=nvptx-none --without-cuda-driver --enable-checking=release --build=x86_64-linux-gnu --host=x86_64-linux-gnu --target=x86_64-linux-gnu > Thread model: posix > gcc version 7.5.0 (Ubuntu 7.5.0-3ubuntu1~18.04) > > Even if this becomes worked-around upstream, I suspect it would be good to mention > this trick in https://sourceware.org/glibc/wiki/Testing/Builds for those > working on older glibc versions. > > Considering that I my build directory is in: > > /home/efficios/git/glibc-build > > and I configure my glibc from the build directory with: > "../glibc/configure --prefix=/home/efficios/glibc-test" > > Where exactly should I copy libgcc_s.so.1 and libstdc++.so.6 ? I tried copying > them to /home/efficios/git/glibc-build/ but it does not appear to fix the issue. hm then the issue may be something else. the build directory should be early in the runtime library search path so if the right libs are there then things should work. > I notice that other .so are pulled into /home/efficios/git/glibc-build/testroot.pristine > and then /home/efficios/git/glibc-build/testroot.root though. Is it where I should > copy them, and if so, under which subdirectory are they expected ? that is used if the test is running in a container (e.g. resolver tests that use custom /etc/resolv.conf) i'd try to run strace -f ./testrun.sh nptl/tst-rseq-nptl --direct 2>strace.log and grep libgcc_s strace.log
----- On Apr 30, 2020, at 6:40 AM, Szabolcs Nagy szabolcs.nagy@arm.com wrote: > The 04/29/2020 08:57, Mathieu Desnoyers wrote: >> ----- On Apr 29, 2020, at 5:01 AM, Florian Weimer fw@deneb.enyo.de wrote: >> >> > * Szabolcs Nagy: >> > >> >> cancel tests work for me on an ubuntu system because >> >> of /etc/ld.so.cache, but that may not be present >> >> or the system may not be glibc based at all. >> > >> > I see. >> > >> >> i always do the cp because i build gcc myself (usually >> >> close to current master) and don't install it to the >> >> system path which means at compile time and runtime >> >> different libraries are used if i dont copy >> > >> > I wonder if we should do this automatically (maybe using a symbolic >> > link), at least of GCC and ld can find these shared objects. Maybe >> > it's possible to tell ld to provide output that makes it easy to >> > locate the actual objects in the file system? >> > >> > (Trimmed the Cc: list.) >> >> For the records, here is the output I get here: >> >> $ gcc --print-file-name libgcc_s.so.1 >> /usr/lib/gcc/x86_64-linux-gnu/7/libgcc_s.so.1 >> $ gcc --print-file-name libstdc++.so.6 >> /usr/lib/gcc/x86_64-linux-gnu/7/../../../x86_64-linux-gnu/libstdc++.so.6 >> $ gcc -v >> Using built-in specs. >> COLLECT_GCC=gcc >> COLLECT_LTO_WRAPPER=/usr/lib/gcc/x86_64-linux-gnu/7/lto-wrapper >> OFFLOAD_TARGET_NAMES=nvptx-none >> OFFLOAD_TARGET_DEFAULT=1 >> Target: x86_64-linux-gnu >> Configured with: ../src/configure -v --with-pkgversion='Ubuntu >> 7.5.0-3ubuntu1~18.04' --with-bugurl=file:///usr/share/doc/gcc-7/README.Bugs >> --enable-languages=c,ada,c++,go,brig,d,fortran,objc,obj-c++ --prefix=/usr >> --with-gcc-major-version-only --program-suffix=-7 >> --program-prefix=x86_64-linux-gnu- --enable-shared --enable-linker-build-id >> --libexecdir=/usr/lib --without-included-gettext --enable-threads=posix >> --libdir=/usr/lib --enable-nls --enable-bootstrap --enable-clocale=gnu >> --enable-libstdcxx-debug --enable-libstdcxx-time=yes >> --with-default-libstdcxx-abi=new --enable-gnu-unique-object >> --disable-vtable-verify --enable-libmpx --enable-plugin --enable-default-pie >> --with-system-zlib --with-target-system-zlib --enable-objc-gc=auto >> --enable-multiarch --disable-werror --with-arch-32=i686 --with-abi=m64 >> --with-multilib-list=m32,m64,mx32 --enable-multilib --with-tune=generic >> --enable-offload-targets=nvptx-none --without-cuda-driver >> --enable-checking=release --build=x86_64-linux-gnu --host=x86_64-linux-gnu >> --target=x86_64-linux-gnu >> Thread model: posix >> gcc version 7.5.0 (Ubuntu 7.5.0-3ubuntu1~18.04) >> >> Even if this becomes worked-around upstream, I suspect it would be good to >> mention >> this trick in https://sourceware.org/glibc/wiki/Testing/Builds for those >> working on older glibc versions. >> >> Considering that I my build directory is in: >> >> /home/efficios/git/glibc-build >> >> and I configure my glibc from the build directory with: >> "../glibc/configure --prefix=/home/efficios/glibc-test" >> >> Where exactly should I copy libgcc_s.so.1 and libstdc++.so.6 ? I tried copying >> them to /home/efficios/git/glibc-build/ but it does not appear to fix the issue. > > hm then the issue may be something else. > > the build directory should be early in the runtime > library search path so if the right libs are there > then things should work. > >> I notice that other .so are pulled into >> /home/efficios/git/glibc-build/testroot.pristine >> and then /home/efficios/git/glibc-build/testroot.root though. Is it where I >> should >> copy them, and if so, under which subdirectory are they expected ? > > that is used if the test is running in a container > (e.g. resolver tests that use custom /etc/resolv.conf) > > i'd try to run > > strace -f ./testrun.sh nptl/tst-rseq-nptl --direct 2>strace.log > > and grep libgcc_s strace.log After a few attempts, it appears that the issue was that I had already run "make check" in my build directory beforehand, so even though I then copied over the .so files to the build directory top level, it did not populate them into the testroot.pristine nor testroot.root directories for the following "make check". If I delete both testroot.pristine and testroot.root, and make check after making sure both .so are in the toplevel build directory, then they are copied over to testroot.pristine, but testroot.root is not created at all (and the test still fails). Starting from a brand new empty build directory, and copying the .so files before the first make check gets the test to run fine! By doing so, the .so are copied into testroot.pristine and testroot.root. Thanks, Mathieu
diff --git a/NEWS b/NEWS index e0379fc53c..cf3e05d5f9 100644 --- a/NEWS +++ b/NEWS @@ -9,7 +9,17 @@ Version 2.32 Major new features: - * New locale added: ckb_IQ (Kurdish/Sorani spoken in Iraq) +* New locale added: ckb_IQ (Kurdish/Sorani spoken in Iraq) + +* Support for automatically registering threads with the Linux rseq(2) + system call has been added. This system call is implemented starting + from Linux 4.18. The Restartable Sequences ABI accelerates user-space + operations on per-cpu data. It allows user-space to perform updates + on per-cpu data without requiring heavy-weight atomic operations. + Automatically registering threads allows all libraries, including libc, + to make immediate use of the rseq(2) support by using the documented ABI. + See 'man 2 rseq' for the details of the ABI shared between libc and the + kernel. Deprecated and removed features, and other changes affecting compatibility: diff --git a/elf/libc_early_init.c b/elf/libc_early_init.c index 1ac66d895d..30466afea0 100644 --- a/elf/libc_early_init.c +++ b/elf/libc_early_init.c @@ -18,10 +18,13 @@ #include <ctype.h> #include <libc-early-init.h> +#include <rseq-internal.h> void __libc_early_init (void) { /* Initialize ctype data. */ __ctype_init (); + /* Register rseq ABI to the kernel. */ + (void) rseq_register_current_thread (); } diff --git a/manual/threads.texi b/manual/threads.texi index 0858ef8f92..59f634e432 100644 --- a/manual/threads.texi +++ b/manual/threads.texi @@ -9,8 +9,10 @@ This chapter describes functions used for managing threads. POSIX threads. @menu -* ISO C Threads:: Threads based on the ISO C specification. -* POSIX Threads:: Threads based on the POSIX specification. +* ISO C Threads:: Threads based on the ISO C specification. +* POSIX Threads:: Threads based on the POSIX specification. +* Restartable Sequences:: Linux-specific Restartable Sequences + integration. @end menu @@ -881,3 +883,27 @@ Behaves like @code{pthread_timedjoin_np} except that the absolute time in @c pthread_spin_unlock @c pthread_testcancel @c pthread_yield + +@node Restartable Sequences +@section Restartable Sequences +@cindex rseq + +This section describes @theglibc{} Restartable Sequences integration. + +@deftypevar {struct rseq} __rseq_abi +@standards{GNU, sys/rseq.h} +@Theglibc{} implements a @code{__rseq_abi} TLS symbol to interact with the +Restartable Sequences system call (Linux-specific). The layout of this +structure is defined by the Linux kernel @file{linux/rseq.h} UAPI. +Registration of each thread's @code{__rseq_abi} is performed by +@theglibc{} at libc initialization and pthread creation. +@end deftypevar + +@deftypevr Macro int RSEQ_SIG +@standards{GNU, sys/rseq.h} +Each supported architecture provide a @code{RSEQ_SIG} macro in +@file{sys/rseq.h} which contains a signature. That signature is expected to be +present in the code before each Restartable Sequences abort handler. Failure +to provide the expected signature may terminate the process with a Segmentation +fault. +@end deftypevr diff --git a/misc/rseq-internal.h b/misc/rseq-internal.h new file mode 100644 index 0000000000..d564cf1bc3 --- /dev/null +++ b/misc/rseq-internal.h @@ -0,0 +1,33 @@ +/* Restartable Sequences internal API. Stub version. + Copyright (C) 2019-2020 Free Software Foundation, Inc. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + <http://www.gnu.org/licenses/>. */ + +#ifndef RSEQ_INTERNAL_H +#define RSEQ_INTERNAL_H + +static inline int +rseq_register_current_thread (void) +{ + return -1; +} + +static inline int +rseq_unregister_current_thread (void) +{ + return -1; +} + +#endif /* rseq-internal.h */ diff --git a/nptl/pthread_create.c b/nptl/pthread_create.c index afd379e89a..1ff248042e 100644 --- a/nptl/pthread_create.c +++ b/nptl/pthread_create.c @@ -33,6 +33,7 @@ #include <default-sched.h> #include <futex-internal.h> #include <tls-setup.h> +#include <rseq-internal.h> #include "libioP.h" #include <shlib-compat.h> @@ -384,6 +385,9 @@ START_THREAD_DEFN /* Initialize pointers to locale data. */ __ctype_init (); + /* Register rseq TLS to the kernel. */ + (void) rseq_register_current_thread (); + #ifndef __ASSUME_SET_ROBUST_LIST if (__set_robust_list_avail >= 0) #endif @@ -578,6 +582,14 @@ START_THREAD_DEFN process is really dead since 'clone' got passed the CLONE_CHILD_CLEARTID flag. The 'tid' field in the TCB will be set to zero. + rseq TLS is still registered at this point. Rely on implicit unregistration + performed by the kernel on thread teardown. This is not a problem because the + rseq TLS lives on the stack, and the stack outlives the thread. If TCB + allocation is ever changed, additional steps may be required, such as + performing explicit rseq unregistration before reclaiming the rseq TLS area + memory. It is NOT sufficient to block signals because the kernel may write + to the rseq area even without signals. + The exit code is zero since in case all threads exit by calling 'pthread_exit' the exit status must be 0 (zero). */ __exit_thread (); diff --git a/sysdeps/unix/sysv/linux/Makefile b/sysdeps/unix/sysv/linux/Makefile index 60dc5cf9e5..6c6f669d21 100644 --- a/sysdeps/unix/sysv/linux/Makefile +++ b/sysdeps/unix/sysv/linux/Makefile @@ -41,7 +41,7 @@ update-syscall-lists: arch-syscall.h endif ifeq ($(subdir),csu) -sysdep_routines += errno-loc +sysdep_routines += errno-loc rseq-sym endif ifeq ($(subdir),assert) @@ -92,7 +92,8 @@ sysdep_headers += sys/mount.h sys/acct.h sys/sysctl.h \ bits/termios-baud.h bits/termios-c_cflag.h \ bits/termios-c_lflag.h bits/termios-tcflow.h \ bits/termios-misc.h \ - bits/ipc-perm.h + bits/ipc-perm.h \ + sys/rseq.h bits/rseq.h tests += tst-clone tst-clone2 tst-clone3 tst-fanotify tst-personality \ tst-quota tst-sync_file_range tst-sysconf-iov_max tst-ttyname \ diff --git a/sysdeps/unix/sysv/linux/Versions b/sysdeps/unix/sysv/linux/Versions index d385085c61..52ca223ab2 100644 --- a/sysdeps/unix/sysv/linux/Versions +++ b/sysdeps/unix/sysv/linux/Versions @@ -177,6 +177,9 @@ libc { GLIBC_2.30 { getdents64; gettid; tgkill; } + GLIBC_2.32 { + __rseq_abi; + } GLIBC_PRIVATE { # functions used in other libraries __syscall_rt_sigqueueinfo; diff --git a/sysdeps/unix/sysv/linux/aarch64/bits/rseq.h b/sysdeps/unix/sysv/linux/aarch64/bits/rseq.h new file mode 100644 index 0000000000..e272c30446 --- /dev/null +++ b/sysdeps/unix/sysv/linux/aarch64/bits/rseq.h @@ -0,0 +1,43 @@ +/* Restartable Sequences Linux aarch64 architecture header. + Copyright (C) 2019-2020 Free Software Foundation, Inc. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + <http://www.gnu.org/licenses/>. */ + +#ifndef _SYS_RSEQ_H +# error "Never use <bits/rseq.h> directly; include <sys/rseq.h> instead." +#endif + +/* RSEQ_SIG is a signature required before each abort handler code. + + It is a 32-bit value that maps to actual architecture code compiled + into applications and libraries. It needs to be defined for each + architecture. When choosing this value, it needs to be taken into + account that generating invalid instructions may have ill effects on + tools like objdump, and may also have impact on the CPU speculative + execution efficiency in some cases. + + aarch64 -mbig-endian generates mixed endianness code vs data: + little-endian code and big-endian data. Ensure the RSEQ_SIG signature + matches code endianness. */ + +#define RSEQ_SIG_CODE 0xd428bc00 /* BRK #0x45E0. */ + +#ifdef __AARCH64EB__ +#define RSEQ_SIG_DATA 0x00bc28d4 /* BRK #0x45E0. */ +#else +#define RSEQ_SIG_DATA RSEQ_SIG_CODE +#endif + +#define RSEQ_SIG RSEQ_SIG_DATA diff --git a/sysdeps/unix/sysv/linux/aarch64/libc.abilist b/sysdeps/unix/sysv/linux/aarch64/libc.abilist index a4c31932cb..fa0702d4ac 100644 --- a/sysdeps/unix/sysv/linux/aarch64/libc.abilist +++ b/sysdeps/unix/sysv/linux/aarch64/libc.abilist @@ -2145,3 +2145,4 @@ GLIBC_2.30 getdents64 F GLIBC_2.30 gettid F GLIBC_2.30 tgkill F GLIBC_2.30 twalk_r F +GLIBC_2.32 __rseq_abi T 0x20 diff --git a/sysdeps/unix/sysv/linux/alpha/libc.abilist b/sysdeps/unix/sysv/linux/alpha/libc.abilist index e7f2174ac2..482c486272 100644 --- a/sysdeps/unix/sysv/linux/alpha/libc.abilist +++ b/sysdeps/unix/sysv/linux/alpha/libc.abilist @@ -2225,6 +2225,7 @@ GLIBC_2.30 getdents64 F GLIBC_2.30 gettid F GLIBC_2.30 tgkill F GLIBC_2.30 twalk_r F +GLIBC_2.32 __rseq_abi T 0x20 GLIBC_2.4 _IO_fprintf F GLIBC_2.4 _IO_printf F GLIBC_2.4 _IO_sprintf F diff --git a/sysdeps/unix/sysv/linux/arm/be/libc.abilist b/sysdeps/unix/sysv/linux/arm/be/libc.abilist index b152c0e24a..753f643c3b 100644 --- a/sysdeps/unix/sysv/linux/arm/be/libc.abilist +++ b/sysdeps/unix/sysv/linux/arm/be/libc.abilist @@ -133,6 +133,7 @@ GLIBC_2.30 twalk_r F GLIBC_2.31 msgctl F GLIBC_2.31 semctl F GLIBC_2.31 shmctl F +GLIBC_2.32 __rseq_abi T 0x20 GLIBC_2.4 _Exit F GLIBC_2.4 _IO_2_1_stderr_ D 0xa0 GLIBC_2.4 _IO_2_1_stdin_ D 0xa0 diff --git a/sysdeps/unix/sysv/linux/arm/bits/rseq.h b/sysdeps/unix/sysv/linux/arm/bits/rseq.h new file mode 100644 index 0000000000..45a2118dbc --- /dev/null +++ b/sysdeps/unix/sysv/linux/arm/bits/rseq.h @@ -0,0 +1,83 @@ +/* Restartable Sequences Linux arm architecture header. + Copyright (C) 2019-2020 Free Software Foundation, Inc. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + <http://www.gnu.org/licenses/>. */ + +#ifndef _SYS_RSEQ_H +# error "Never use <bits/rseq.h> directly; include <sys/rseq.h> instead." +#endif + +/* + RSEQ_SIG is a signature required before each abort handler code. + + It is a 32-bit value that maps to actual architecture code compiled + into applications and libraries. It needs to be defined for each + architecture. When choosing this value, it needs to be taken into + account that generating invalid instructions may have ill effects on + tools like objdump, and may also have impact on the CPU speculative + execution efficiency in some cases. + + - ARM little endian + + RSEQ_SIG uses the udf A32 instruction with an uncommon immediate operand + value 0x5de3. This traps if user-space reaches this instruction by mistake, + and the uncommon operand ensures the kernel does not move the instruction + pointer to attacker-controlled code on rseq abort. + + The instruction pattern in the A32 instruction set is: + + e7f5def3 udf #24035 ; 0x5de3 + + This translates to the following instruction pattern in the T16 instruction + set: + + little endian: + def3 udf #243 ; 0xf3 + e7f5 b.n <7f5> + + - ARMv6+ big endian (BE8): + + ARMv6+ -mbig-endian generates mixed endianness code vs data: little-endian + code and big-endian data. The data value of the signature needs to have its + byte order reversed to generate the trap instruction: + + Data: 0xf3def5e7 + + Translates to this A32 instruction pattern: + + e7f5def3 udf #24035 ; 0x5de3 + + Translates to this T16 instruction pattern: + + def3 udf #243 ; 0xf3 + e7f5 b.n <7f5> + + - Prior to ARMv6 big endian (BE32): + + Prior to ARMv6, -mbig-endian generates big-endian code and data + (which match), so the endianness of the data representation of the + signature should not be reversed. However, the choice between BE32 + and BE8 is done by the linker, so we cannot know whether code and + data endianness will be mixed before the linker is invoked. So rather + than try to play tricks with the linker, the rseq signature is simply + data (not a trap instruction) prior to ARMv6 on big endian. This is + why the signature is expressed as data (.word) rather than as + instruction (.inst) in assembler. */ + +#ifdef __ARMEB__ +#define RSEQ_SIG 0xf3def5e7 /* udf #24035 ; 0x5de3 (ARMv6+) */ +#else +#define RSEQ_SIG 0xe7f5def3 /* udf #24035 ; 0x5de3 */ +#endif diff --git a/sysdeps/unix/sysv/linux/arm/le/libc.abilist b/sysdeps/unix/sysv/linux/arm/le/libc.abilist index 9371927927..97b1081fa5 100644 --- a/sysdeps/unix/sysv/linux/arm/le/libc.abilist +++ b/sysdeps/unix/sysv/linux/arm/le/libc.abilist @@ -130,6 +130,7 @@ GLIBC_2.30 getdents64 F GLIBC_2.30 gettid F GLIBC_2.30 tgkill F GLIBC_2.30 twalk_r F +GLIBC_2.32 __rseq_abi T 0x20 GLIBC_2.4 _Exit F GLIBC_2.4 _IO_2_1_stderr_ D 0xa0 GLIBC_2.4 _IO_2_1_stdin_ D 0xa0 diff --git a/sysdeps/unix/sysv/linux/bits/rseq.h b/sysdeps/unix/sysv/linux/bits/rseq.h new file mode 100644 index 0000000000..b8a63ed26d --- /dev/null +++ b/sysdeps/unix/sysv/linux/bits/rseq.h @@ -0,0 +1,29 @@ +/* Restartable Sequences architecture header. Stub version. + Copyright (C) 2019-2020 Free Software Foundation, Inc. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + <http://www.gnu.org/licenses/>. */ + +#ifndef _SYS_RSEQ_H +# error "Never use <bits/rseq.h> directly; include <sys/rseq.h> instead." +#endif + +/* RSEQ_SIG is a signature required before each abort handler code. + + It is a 32-bit value that maps to actual architecture code compiled + into applications and libraries. It needs to be defined for each + architecture. When choosing this value, it needs to be taken into + account that generating invalid instructions may have ill effects on + tools like objdump, and may also have impact on the CPU speculative + execution efficiency in some cases. */ diff --git a/sysdeps/unix/sysv/linux/csky/libc.abilist b/sysdeps/unix/sysv/linux/csky/libc.abilist index 9b3cee65bb..ed94510f23 100644 --- a/sysdeps/unix/sysv/linux/csky/libc.abilist +++ b/sysdeps/unix/sysv/linux/csky/libc.abilist @@ -2089,3 +2089,4 @@ GLIBC_2.30 getdents64 F GLIBC_2.30 gettid F GLIBC_2.30 tgkill F GLIBC_2.30 twalk_r F +GLIBC_2.32 __rseq_abi T 0x20 diff --git a/sysdeps/unix/sysv/linux/hppa/libc.abilist b/sysdeps/unix/sysv/linux/hppa/libc.abilist index df6d96fbae..3d60443799 100644 --- a/sysdeps/unix/sysv/linux/hppa/libc.abilist +++ b/sysdeps/unix/sysv/linux/hppa/libc.abilist @@ -2046,6 +2046,7 @@ GLIBC_2.30 getdents64 F GLIBC_2.30 gettid F GLIBC_2.30 tgkill F GLIBC_2.30 twalk_r F +GLIBC_2.32 __rseq_abi T 0x20 GLIBC_2.4 __confstr_chk F GLIBC_2.4 __fgets_chk F GLIBC_2.4 __fgets_unlocked_chk F diff --git a/sysdeps/unix/sysv/linux/i386/libc.abilist b/sysdeps/unix/sysv/linux/i386/libc.abilist index fcb625b6bf..15f0baa54f 100644 --- a/sysdeps/unix/sysv/linux/i386/libc.abilist +++ b/sysdeps/unix/sysv/linux/i386/libc.abilist @@ -2212,6 +2212,7 @@ GLIBC_2.30 getdents64 F GLIBC_2.30 gettid F GLIBC_2.30 tgkill F GLIBC_2.30 twalk_r F +GLIBC_2.32 __rseq_abi T 0x20 GLIBC_2.4 __confstr_chk F GLIBC_2.4 __fgets_chk F GLIBC_2.4 __fgets_unlocked_chk F diff --git a/sysdeps/unix/sysv/linux/ia64/libc.abilist b/sysdeps/unix/sysv/linux/ia64/libc.abilist index cb556c5998..8944cb2cc8 100644 --- a/sysdeps/unix/sysv/linux/ia64/libc.abilist +++ b/sysdeps/unix/sysv/linux/ia64/libc.abilist @@ -2078,6 +2078,7 @@ GLIBC_2.30 getdents64 F GLIBC_2.30 gettid F GLIBC_2.30 tgkill F GLIBC_2.30 twalk_r F +GLIBC_2.32 __rseq_abi T 0x20 GLIBC_2.4 __confstr_chk F GLIBC_2.4 __fgets_chk F GLIBC_2.4 __fgets_unlocked_chk F diff --git a/sysdeps/unix/sysv/linux/m68k/coldfire/libc.abilist b/sysdeps/unix/sysv/linux/m68k/coldfire/libc.abilist index 5e3cdea246..ae100260b9 100644 --- a/sysdeps/unix/sysv/linux/m68k/coldfire/libc.abilist +++ b/sysdeps/unix/sysv/linux/m68k/coldfire/libc.abilist @@ -134,6 +134,7 @@ GLIBC_2.30 twalk_r F GLIBC_2.31 msgctl F GLIBC_2.31 semctl F GLIBC_2.31 shmctl F +GLIBC_2.32 __rseq_abi T 0x20 GLIBC_2.4 _Exit F GLIBC_2.4 _IO_2_1_stderr_ D 0x98 GLIBC_2.4 _IO_2_1_stdin_ D 0x98 diff --git a/sysdeps/unix/sysv/linux/m68k/m680x0/libc.abilist b/sysdeps/unix/sysv/linux/m68k/m680x0/libc.abilist index ea5e7a41af..c2b3f91d17 100644 --- a/sysdeps/unix/sysv/linux/m68k/m680x0/libc.abilist +++ b/sysdeps/unix/sysv/linux/m68k/m680x0/libc.abilist @@ -2158,6 +2158,7 @@ GLIBC_2.30 twalk_r F GLIBC_2.31 msgctl F GLIBC_2.31 semctl F GLIBC_2.31 shmctl F +GLIBC_2.32 __rseq_abi T 0x20 GLIBC_2.4 __confstr_chk F GLIBC_2.4 __fgets_chk F GLIBC_2.4 __fgets_unlocked_chk F diff --git a/sysdeps/unix/sysv/linux/microblaze/be/libc.abilist b/sysdeps/unix/sysv/linux/microblaze/be/libc.abilist index ac55b0acd7..ba9ebeae22 100644 --- a/sysdeps/unix/sysv/linux/microblaze/be/libc.abilist +++ b/sysdeps/unix/sysv/linux/microblaze/be/libc.abilist @@ -2140,3 +2140,4 @@ GLIBC_2.30 twalk_r F GLIBC_2.31 msgctl F GLIBC_2.31 semctl F GLIBC_2.31 shmctl F +GLIBC_2.32 __rseq_abi T 0x20 diff --git a/sysdeps/unix/sysv/linux/microblaze/le/libc.abilist b/sysdeps/unix/sysv/linux/microblaze/le/libc.abilist index f7ced487f7..a1f4f9a1c2 100644 --- a/sysdeps/unix/sysv/linux/microblaze/le/libc.abilist +++ b/sysdeps/unix/sysv/linux/microblaze/le/libc.abilist @@ -2137,3 +2137,4 @@ GLIBC_2.30 getdents64 F GLIBC_2.30 gettid F GLIBC_2.30 tgkill F GLIBC_2.30 twalk_r F +GLIBC_2.32 __rseq_abi T 0x20 diff --git a/sysdeps/unix/sysv/linux/mips/bits/rseq.h b/sysdeps/unix/sysv/linux/mips/bits/rseq.h new file mode 100644 index 0000000000..4eee59265a --- /dev/null +++ b/sysdeps/unix/sysv/linux/mips/bits/rseq.h @@ -0,0 +1,62 @@ +/* Restartable Sequences Linux mips architecture header. + Copyright (C) 2019-2020 Free Software Foundation, Inc. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + <http://www.gnu.org/licenses/>. */ + +#ifndef _SYS_RSEQ_H +# error "Never use <bits/rseq.h> directly; include <sys/rseq.h> instead." +#endif + +/* RSEQ_SIG is a signature required before each abort handler code. + + It is a 32-bit value that maps to actual architecture code compiled + into applications and libraries. It needs to be defined for each + architecture. When choosing this value, it needs to be taken into + account that generating invalid instructions may have ill effects on + tools like objdump, and may also have impact on the CPU speculative + execution efficiency in some cases. + + RSEQ_SIG uses the break instruction. The instruction pattern is: + + On MIPS: + 0350000d break 0x350 + + On nanoMIPS: + 00100350 break 0x350 + + On microMIPS: + 0000d407 break 0x350 + + For nanoMIPS32 and microMIPS, the instruction stream is encoded as + 16-bit halfwords, so the signature halfwords need to be swapped + accordingly for little-endian. */ + +#if defined(__nanomips__) +# ifdef __MIPSEL__ +# define RSEQ_SIG 0x03500010 +# else +# define RSEQ_SIG 0x00100350 +# endif +#elif defined(__mips_micromips) +# ifdef __MIPSEL__ +# define RSEQ_SIG 0xd4070000 +# else +# define RSEQ_SIG 0x0000d407 +# endif +#elif defined(__mips__) +# define RSEQ_SIG 0x0350000d +#else +/* Unknown MIPS architecture. */ +#endif diff --git a/sysdeps/unix/sysv/linux/mips/mips32/fpu/libc.abilist b/sysdeps/unix/sysv/linux/mips/mips32/fpu/libc.abilist index 06c2e64edd..333229c27a 100644 --- a/sysdeps/unix/sysv/linux/mips/mips32/fpu/libc.abilist +++ b/sysdeps/unix/sysv/linux/mips/mips32/fpu/libc.abilist @@ -2129,6 +2129,7 @@ GLIBC_2.30 getdents64 F GLIBC_2.30 gettid F GLIBC_2.30 tgkill F GLIBC_2.30 twalk_r F +GLIBC_2.32 __rseq_abi T 0x20 GLIBC_2.4 __confstr_chk F GLIBC_2.4 __fgets_chk F GLIBC_2.4 __fgets_unlocked_chk F diff --git a/sysdeps/unix/sysv/linux/mips/mips32/nofpu/libc.abilist b/sysdeps/unix/sysv/linux/mips/mips32/nofpu/libc.abilist index bdfd073b86..17772e9a40 100644 --- a/sysdeps/unix/sysv/linux/mips/mips32/nofpu/libc.abilist +++ b/sysdeps/unix/sysv/linux/mips/mips32/nofpu/libc.abilist @@ -2127,6 +2127,7 @@ GLIBC_2.30 getdents64 F GLIBC_2.30 gettid F GLIBC_2.30 tgkill F GLIBC_2.30 twalk_r F +GLIBC_2.32 __rseq_abi T 0x20 GLIBC_2.4 __confstr_chk F GLIBC_2.4 __fgets_chk F GLIBC_2.4 __fgets_unlocked_chk F diff --git a/sysdeps/unix/sysv/linux/mips/mips64/n32/libc.abilist b/sysdeps/unix/sysv/linux/mips/mips64/n32/libc.abilist index 3d61d4974a..497203cd44 100644 --- a/sysdeps/unix/sysv/linux/mips/mips64/n32/libc.abilist +++ b/sysdeps/unix/sysv/linux/mips/mips64/n32/libc.abilist @@ -2135,6 +2135,7 @@ GLIBC_2.30 getdents64 F GLIBC_2.30 gettid F GLIBC_2.30 tgkill F GLIBC_2.30 twalk_r F +GLIBC_2.32 __rseq_abi T 0x20 GLIBC_2.4 __confstr_chk F GLIBC_2.4 __fgets_chk F GLIBC_2.4 __fgets_unlocked_chk F diff --git a/sysdeps/unix/sysv/linux/mips/mips64/n64/libc.abilist b/sysdeps/unix/sysv/linux/mips/mips64/n64/libc.abilist index 675acca5db..f5f48ce5c6 100644 --- a/sysdeps/unix/sysv/linux/mips/mips64/n64/libc.abilist +++ b/sysdeps/unix/sysv/linux/mips/mips64/n64/libc.abilist @@ -2129,6 +2129,7 @@ GLIBC_2.30 getdents64 F GLIBC_2.30 gettid F GLIBC_2.30 tgkill F GLIBC_2.30 twalk_r F +GLIBC_2.32 __rseq_abi T 0x20 GLIBC_2.4 __confstr_chk F GLIBC_2.4 __fgets_chk F GLIBC_2.4 __fgets_unlocked_chk F diff --git a/sysdeps/unix/sysv/linux/nios2/libc.abilist b/sysdeps/unix/sysv/linux/nios2/libc.abilist index 7fec0c9670..4152f4f734 100644 --- a/sysdeps/unix/sysv/linux/nios2/libc.abilist +++ b/sysdeps/unix/sysv/linux/nios2/libc.abilist @@ -2178,3 +2178,4 @@ GLIBC_2.30 getdents64 F GLIBC_2.30 gettid F GLIBC_2.30 tgkill F GLIBC_2.30 twalk_r F +GLIBC_2.32 __rseq_abi T 0x20 diff --git a/sysdeps/unix/sysv/linux/powerpc/bits/rseq.h b/sysdeps/unix/sysv/linux/powerpc/bits/rseq.h new file mode 100644 index 0000000000..9d10000a6e --- /dev/null +++ b/sysdeps/unix/sysv/linux/powerpc/bits/rseq.h @@ -0,0 +1,37 @@ +/* Restartable Sequences Linux powerpc architecture header. + Copyright (C) 2019-2020 Free Software Foundation, Inc. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + <http://www.gnu.org/licenses/>. */ + +#ifndef _SYS_RSEQ_H +# error "Never use <bits/rseq.h> directly; include <sys/rseq.h> instead." +#endif + +/* RSEQ_SIG is a signature required before each abort handler code. + + It is a 32-bit value that maps to actual architecture code compiled + into applications and libraries. It needs to be defined for each + architecture. When choosing this value, it needs to be taken into + account that generating invalid instructions may have ill effects on + tools like objdump, and may also have impact on the CPU speculative + execution efficiency in some cases. + + RSEQ_SIG uses the following trap instruction: + + powerpc-be: 0f e5 00 0b twui r5,11 + powerpc64-le: 0b 00 e5 0f twui r5,11 + powerpc64-be: 0f e5 00 0b twui r5,11 */ + +#define RSEQ_SIG 0x0fe5000b diff --git a/sysdeps/unix/sysv/linux/powerpc/powerpc32/fpu/libc.abilist b/sysdeps/unix/sysv/linux/powerpc/powerpc32/fpu/libc.abilist index 1e8ff6f83e..210d1795c1 100644 --- a/sysdeps/unix/sysv/linux/powerpc/powerpc32/fpu/libc.abilist +++ b/sysdeps/unix/sysv/linux/powerpc/powerpc32/fpu/libc.abilist @@ -2185,6 +2185,7 @@ GLIBC_2.30 getdents64 F GLIBC_2.30 gettid F GLIBC_2.30 tgkill F GLIBC_2.30 twalk_r F +GLIBC_2.32 __rseq_abi T 0x20 GLIBC_2.4 _IO_fprintf F GLIBC_2.4 _IO_printf F GLIBC_2.4 _IO_sprintf F diff --git a/sysdeps/unix/sysv/linux/powerpc/powerpc32/nofpu/libc.abilist b/sysdeps/unix/sysv/linux/powerpc/powerpc32/nofpu/libc.abilist index b5a0751d90..6a3cde5fb6 100644 --- a/sysdeps/unix/sysv/linux/powerpc/powerpc32/nofpu/libc.abilist +++ b/sysdeps/unix/sysv/linux/powerpc/powerpc32/nofpu/libc.abilist @@ -2218,6 +2218,7 @@ GLIBC_2.30 getdents64 F GLIBC_2.30 gettid F GLIBC_2.30 tgkill F GLIBC_2.30 twalk_r F +GLIBC_2.32 __rseq_abi T 0x20 GLIBC_2.4 _IO_fprintf F GLIBC_2.4 _IO_printf F GLIBC_2.4 _IO_sprintf F diff --git a/sysdeps/unix/sysv/linux/powerpc/powerpc64/be/libc.abilist b/sysdeps/unix/sysv/linux/powerpc/powerpc64/be/libc.abilist index 0c86217fc6..4f243acb2a 100644 --- a/sysdeps/unix/sysv/linux/powerpc/powerpc64/be/libc.abilist +++ b/sysdeps/unix/sysv/linux/powerpc/powerpc64/be/libc.abilist @@ -2048,6 +2048,7 @@ GLIBC_2.30 getdents64 F GLIBC_2.30 gettid F GLIBC_2.30 tgkill F GLIBC_2.30 twalk_r F +GLIBC_2.32 __rseq_abi T 0x20 GLIBC_2.4 _IO_fprintf F GLIBC_2.4 _IO_printf F GLIBC_2.4 _IO_sprintf F diff --git a/sysdeps/unix/sysv/linux/powerpc/powerpc64/le/libc.abilist b/sysdeps/unix/sysv/linux/powerpc/powerpc64/le/libc.abilist index 2229a1dcc0..99d836bc11 100644 --- a/sysdeps/unix/sysv/linux/powerpc/powerpc64/le/libc.abilist +++ b/sysdeps/unix/sysv/linux/powerpc/powerpc64/le/libc.abilist @@ -2247,3 +2247,4 @@ GLIBC_2.30 getdents64 F GLIBC_2.30 gettid F GLIBC_2.30 tgkill F GLIBC_2.30 twalk_r F +GLIBC_2.32 __rseq_abi T 0x20 diff --git a/sysdeps/unix/sysv/linux/riscv/rv64/libc.abilist b/sysdeps/unix/sysv/linux/riscv/rv64/libc.abilist index 31010e6cf7..7694392cd5 100644 --- a/sysdeps/unix/sysv/linux/riscv/rv64/libc.abilist +++ b/sysdeps/unix/sysv/linux/riscv/rv64/libc.abilist @@ -2107,3 +2107,4 @@ GLIBC_2.30 getdents64 F GLIBC_2.30 gettid F GLIBC_2.30 tgkill F GLIBC_2.30 twalk_r F +GLIBC_2.32 __rseq_abi T 0x20 diff --git a/sysdeps/unix/sysv/linux/rseq-internal.h b/sysdeps/unix/sysv/linux/rseq-internal.h new file mode 100644 index 0000000000..5f7f02f1ec --- /dev/null +++ b/sysdeps/unix/sysv/linux/rseq-internal.h @@ -0,0 +1,73 @@ +/* Restartable Sequences internal API. Linux implementation. + Copyright (C) 2019-2020 Free Software Foundation, Inc. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + <http://www.gnu.org/licenses/>. */ + +#ifndef RSEQ_INTERNAL_H +#define RSEQ_INTERNAL_H + +#include <sysdep.h> +#include <errno.h> +#include <kernel-features.h> +#include <sys/rseq.h> + +#ifdef RSEQ_SIG + +static inline int +rseq_register_current_thread (void) +{ + int rc, ret = 0; + + if (__rseq_abi.cpu_id == RSEQ_CPU_ID_REGISTRATION_FAILED) + return -1; + rc = INTERNAL_SYSCALL_CALL (rseq, &__rseq_abi, sizeof (struct rseq), + 0, RSEQ_SIG); + if (!rc) + goto end; + if (INTERNAL_SYSCALL_ERRNO (rc) != EBUSY) + __rseq_abi.cpu_id = RSEQ_CPU_ID_REGISTRATION_FAILED; + ret = -1; +end: + return ret; +} + +static inline int +rseq_unregister_current_thread (void) +{ + int rc, ret = 0; + + rc = INTERNAL_SYSCALL_CALL (rseq, &__rseq_abi, sizeof (struct rseq), + RSEQ_FLAG_UNREGISTER, RSEQ_SIG); + if (!rc) + goto end; + ret = -1; +end: + return ret; +} +#else +static inline int +rseq_register_current_thread (void) +{ + return -1; +} + +static inline int +rseq_unregister_current_thread (void) +{ + return -1; +} +#endif + +#endif /* rseq-internal.h */ diff --git a/sysdeps/unix/sysv/linux/rseq-sym.c b/sysdeps/unix/sysv/linux/rseq-sym.c new file mode 100644 index 0000000000..0e33fab278 --- /dev/null +++ b/sysdeps/unix/sysv/linux/rseq-sym.c @@ -0,0 +1,25 @@ +/* Restartable Sequences exported symbols. Linux Implementation. + Copyright (C) 2019-2020 Free Software Foundation, Inc. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + <http://www.gnu.org/licenses/>. */ + +#include <sys/syscall.h> +#include <stdint.h> +#include <kernel-features.h> +#include <sys/rseq.h> + +__thread struct rseq __rseq_abi = { + .cpu_id = RSEQ_CPU_ID_UNINITIALIZED, +}; diff --git a/sysdeps/unix/sysv/linux/s390/bits/rseq.h b/sysdeps/unix/sysv/linux/s390/bits/rseq.h new file mode 100644 index 0000000000..c25ee67ee7 --- /dev/null +++ b/sysdeps/unix/sysv/linux/s390/bits/rseq.h @@ -0,0 +1,37 @@ +/* Restartable Sequences Linux s390 architecture header. + Copyright (C) 2019-2020 Free Software Foundation, Inc. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + <http://www.gnu.org/licenses/>. */ + +#ifndef _SYS_RSEQ_H +# error "Never use <bits/rseq.h> directly; include <sys/rseq.h> instead." +#endif + +/* RSEQ_SIG is a signature required before each abort handler code. + + It is a 32-bit value that maps to actual architecture code compiled + into applications and libraries. It needs to be defined for each + architecture. When choosing this value, it needs to be taken into + account that generating invalid instructions may have ill effects on + tools like objdump, and may also have impact on the CPU speculative + execution efficiency in some cases. + + RSEQ_SIG uses the trap4 instruction. As Linux does not make use of the + access-register mode nor the linkage stack this instruction will always + cause a special-operation exception (the trap-enabled bit in the DUCT + is and will stay 0). The instruction pattern is + b2 ff 0f ff trap4 4095(%r0) */ + +#define RSEQ_SIG 0xB2FF0FFF diff --git a/sysdeps/unix/sysv/linux/s390/s390-32/libc.abilist b/sysdeps/unix/sysv/linux/s390/s390-32/libc.abilist index 4feca641b0..eb49d11cd1 100644 --- a/sysdeps/unix/sysv/linux/s390/s390-32/libc.abilist +++ b/sysdeps/unix/sysv/linux/s390/s390-32/libc.abilist @@ -2183,6 +2183,7 @@ GLIBC_2.30 twalk_r F GLIBC_2.31 msgctl F GLIBC_2.31 semctl F GLIBC_2.31 shmctl F +GLIBC_2.32 __rseq_abi T 0x20 GLIBC_2.4 _IO_fprintf F GLIBC_2.4 _IO_printf F GLIBC_2.4 _IO_sprintf F diff --git a/sysdeps/unix/sysv/linux/s390/s390-64/libc.abilist b/sysdeps/unix/sysv/linux/s390/s390-64/libc.abilist index efe588a072..442f7e33a8 100644 --- a/sysdeps/unix/sysv/linux/s390/s390-64/libc.abilist +++ b/sysdeps/unix/sysv/linux/s390/s390-64/libc.abilist @@ -2084,6 +2084,7 @@ GLIBC_2.30 getdents64 F GLIBC_2.30 gettid F GLIBC_2.30 tgkill F GLIBC_2.30 twalk_r F +GLIBC_2.32 __rseq_abi T 0x20 GLIBC_2.4 _IO_fprintf F GLIBC_2.4 _IO_printf F GLIBC_2.4 _IO_sprintf F diff --git a/sysdeps/unix/sysv/linux/sh/be/libc.abilist b/sysdeps/unix/sysv/linux/sh/be/libc.abilist index 6bfc2b7439..f47ec14735 100644 --- a/sysdeps/unix/sysv/linux/sh/be/libc.abilist +++ b/sysdeps/unix/sysv/linux/sh/be/libc.abilist @@ -2053,6 +2053,7 @@ GLIBC_2.30 twalk_r F GLIBC_2.31 msgctl F GLIBC_2.31 semctl F GLIBC_2.31 shmctl F +GLIBC_2.32 __rseq_abi T 0x20 GLIBC_2.4 __confstr_chk F GLIBC_2.4 __fgets_chk F GLIBC_2.4 __fgets_unlocked_chk F diff --git a/sysdeps/unix/sysv/linux/sh/le/libc.abilist b/sysdeps/unix/sysv/linux/sh/le/libc.abilist index 4b057bf4a2..8b4557d325 100644 --- a/sysdeps/unix/sysv/linux/sh/le/libc.abilist +++ b/sysdeps/unix/sysv/linux/sh/le/libc.abilist @@ -2050,6 +2050,7 @@ GLIBC_2.30 getdents64 F GLIBC_2.30 gettid F GLIBC_2.30 tgkill F GLIBC_2.30 twalk_r F +GLIBC_2.32 __rseq_abi T 0x20 GLIBC_2.4 __confstr_chk F GLIBC_2.4 __fgets_chk F GLIBC_2.4 __fgets_unlocked_chk F diff --git a/sysdeps/unix/sysv/linux/sparc/sparc32/libc.abilist b/sysdeps/unix/sysv/linux/sparc/sparc32/libc.abilist index 49cd597fd6..52f951c4fe 100644 --- a/sysdeps/unix/sysv/linux/sparc/sparc32/libc.abilist +++ b/sysdeps/unix/sysv/linux/sparc/sparc32/libc.abilist @@ -2174,6 +2174,7 @@ GLIBC_2.30 getdents64 F GLIBC_2.30 gettid F GLIBC_2.30 tgkill F GLIBC_2.30 twalk_r F +GLIBC_2.32 __rseq_abi T 0x20 GLIBC_2.4 _IO_fprintf F GLIBC_2.4 _IO_printf F GLIBC_2.4 _IO_sprintf F diff --git a/sysdeps/unix/sysv/linux/sparc/sparc64/libc.abilist b/sysdeps/unix/sysv/linux/sparc/sparc64/libc.abilist index 95e68e0ba1..2b59f52de0 100644 --- a/sysdeps/unix/sysv/linux/sparc/sparc64/libc.abilist +++ b/sysdeps/unix/sysv/linux/sparc/sparc64/libc.abilist @@ -2101,6 +2101,7 @@ GLIBC_2.30 getdents64 F GLIBC_2.30 gettid F GLIBC_2.30 tgkill F GLIBC_2.30 twalk_r F +GLIBC_2.32 __rseq_abi T 0x20 GLIBC_2.4 __confstr_chk F GLIBC_2.4 __fgets_chk F GLIBC_2.4 __fgets_unlocked_chk F diff --git a/sysdeps/unix/sysv/linux/sys/rseq.h b/sysdeps/unix/sysv/linux/sys/rseq.h new file mode 100644 index 0000000000..503dce4cac --- /dev/null +++ b/sysdeps/unix/sysv/linux/sys/rseq.h @@ -0,0 +1,186 @@ +/* Restartable Sequences exported symbols. Linux header. + Copyright (C) 2019-2020 Free Software Foundation, Inc. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + <http://www.gnu.org/licenses/>. */ + +#ifndef _SYS_RSEQ_H +#define _SYS_RSEQ_H 1 + +/* Architecture-specific rseq signature. */ +#include <bits/rseq.h> +#include <stdint.h> +#include <sys/cdefs.h> + +#ifdef __has_include +# if __has_include ("linux/rseq.h") +# define __GLIBC_HAVE_KERNEL_RSEQ +# endif +#else +# include <linux/version.h> +# if LINUX_VERSION_CODE >= KERNEL_VERSION (4, 18, 0) +# define __GLIBC_HAVE_KERNEL_RSEQ +# endif +#endif + +#ifdef __GLIBC_HAVE_KERNEL_RSEQ +/* We use the structures declarations from the kernel headers. */ +# include <linux/rseq.h> +#else +/* We use a copy of the include/uapi/linux/rseq.h kernel header. */ + +#include <asm/byteorder.h> + +enum rseq_cpu_id_state + { + RSEQ_CPU_ID_UNINITIALIZED = -1, + RSEQ_CPU_ID_REGISTRATION_FAILED = -2, + }; + +enum rseq_flags + { + RSEQ_FLAG_UNREGISTER = (1 << 0), + }; + +enum rseq_cs_flags_bit + { + RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT = 0, + RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT = 1, + RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT = 2, + }; + +enum rseq_cs_flags + { + RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT = + (1U << RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT_BIT), + RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL = + (1U << RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL_BIT), + RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE = + (1U << RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE_BIT), + }; + +/* struct rseq_cs is aligned on 4 * 8 bytes to ensure it is always + contained within a single cache-line. It is usually declared as + link-time constant data. */ +struct rseq_cs + { + /* Version of this structure. */ + uint32_t version; + /* enum rseq_cs_flags. */ + uint32_t flags; + uint64_t start_ip; + /* Offset from start_ip. */ + uint64_t post_commit_offset; + uint64_t abort_ip; +} __attribute__((aligned(4 * sizeof(uint64_t)))); + +/* struct rseq is aligned on 4 * 8 bytes to ensure it is always + contained within a single cache-line. + + A single struct rseq per thread is allowed. */ +struct rseq + { + /* Restartable sequences cpu_id_start field. Updated by the + kernel. Read by user-space with single-copy atomicity + semantics. This field should only be read by the thread which + registered this data structure. Aligned on 32-bit. Always + contains a value in the range of possible CPUs, although the + value may not be the actual current CPU (e.g. if rseq is not + initialized). This CPU number value should always be compared + against the value of the cpu_id field before performing a rseq + commit or returning a value read from a data structure indexed + using the cpu_id_start value. */ + uint32_t cpu_id_start; + /* Restartable sequences cpu_id field. Updated by the kernel. + Read by user-space with single-copy atomicity semantics. This + field should only be read by the thread which registered this + data structure. Aligned on 32-bit. Values + RSEQ_CPU_ID_UNINITIALIZED and RSEQ_CPU_ID_REGISTRATION_FAILED + have a special semantic: the former means "rseq uninitialized", + and latter means "rseq initialization failed". This value is + meant to be read within rseq critical sections and compared + with the cpu_id_start value previously read, before performing + the commit instruction, or read and compared with the + cpu_id_start value before returning a value loaded from a data + structure indexed using the cpu_id_start value. */ + uint32_t cpu_id; + /* Restartable sequences rseq_cs field. + + Contains NULL when no critical section is active for the current + thread, or holds a pointer to the currently active struct rseq_cs. + + Updated by user-space, which sets the address of the currently + active rseq_cs at the beginning of assembly instruction sequence + block, and set to NULL by the kernel when it restarts an assembly + instruction sequence block, as well as when the kernel detects that + it is preempting or delivering a signal outside of the range + targeted by the rseq_cs. Also needs to be set to NULL by user-space + before reclaiming memory that contains the targeted struct rseq_cs. + + Read and set by the kernel. Set by user-space with single-copy + atomicity semantics. This field should only be updated by the + thread which registered this data structure. Aligned on 64-bit. */ + union { + uint64_t ptr64; +#ifdef __LP64__ + uint64_t ptr; +#else + struct { +#if (defined(__BYTE_ORDER) && (__BYTE_ORDER == __BIG_ENDIAN)) || defined(__BIG_ENDIAN) + uint32_t padding; /* Initialized to zero. */ + uint32_t ptr32; +#else /* LITTLE */ + uint32_t ptr32; + uint32_t padding; /* Initialized to zero. */ +#endif /* ENDIAN */ + } ptr; +#endif + } rseq_cs; + + /* Restartable sequences flags field. + + This field should only be updated by the thread which + registered this data structure. Read by the kernel. + Mainly used for single-stepping through rseq critical sections + with debuggers. + + - RSEQ_CS_FLAG_NO_RESTART_ON_PREEMPT + Inhibit instruction sequence block restart on preemption + for this thread. + - RSEQ_CS_FLAG_NO_RESTART_ON_SIGNAL + Inhibit instruction sequence block restart on signal + delivery for this thread. + - RSEQ_CS_FLAG_NO_RESTART_ON_MIGRATE + Inhibit instruction sequence block restart on migration for + this thread. */ + uint32_t flags; + } __attribute__((aligned(4 * sizeof(uint64_t)))); + +#endif + +/* Ensure the compiler supports __attribute__ ((aligned)). */ +_Static_assert (__alignof__ (struct rseq_cs) >= 4 * sizeof(uint64_t), + "alignment"); +_Static_assert (__alignof__ (struct rseq) >= 4 * sizeof(uint64_t), + "alignment"); + +/* Allocations of struct rseq and struct rseq_cs on the heap need to + be aligned on 32 bytes. Therefore, use of malloc is discouraged + because it does not guarantee alignment. posix_memalign should be + used instead. */ + +extern __thread struct rseq __rseq_abi +__attribute__ ((tls_model ("initial-exec"))); + +#endif /* sys/rseq.h */ diff --git a/sysdeps/unix/sysv/linux/x86/bits/rseq.h b/sysdeps/unix/sysv/linux/x86/bits/rseq.h new file mode 100644 index 0000000000..75f52d9788 --- /dev/null +++ b/sysdeps/unix/sysv/linux/x86/bits/rseq.h @@ -0,0 +1,30 @@ +/* Restartable Sequences Linux x86 architecture header. + Copyright (C) 2019-2020 Free Software Foundation, Inc. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + <http://www.gnu.org/licenses/>. */ + +#ifndef _SYS_RSEQ_H +# error "Never use <bits/rseq.h> directly; include <sys/rseq.h> instead." +#endif + +/* RSEQ_SIG is a signature required before each abort handler code. + + RSEQ_SIG is used with the following reserved undefined instructions, which + trap in user-space: + + x86-32: 0f b9 3d 53 30 05 53 ud1 0x53053053,%edi + x86-64: 0f b9 3d 53 30 05 53 ud1 0x53053053(%rip),%edi */ + +#define RSEQ_SIG 0x53053053 diff --git a/sysdeps/unix/sysv/linux/x86_64/64/libc.abilist b/sysdeps/unix/sysv/linux/x86_64/64/libc.abilist index 1f2dbd1451..dfa3fe85ef 100644 --- a/sysdeps/unix/sysv/linux/x86_64/64/libc.abilist +++ b/sysdeps/unix/sysv/linux/x86_64/64/libc.abilist @@ -2059,6 +2059,7 @@ GLIBC_2.30 getdents64 F GLIBC_2.30 gettid F GLIBC_2.30 tgkill F GLIBC_2.30 twalk_r F +GLIBC_2.32 __rseq_abi T 0x20 GLIBC_2.4 __confstr_chk F GLIBC_2.4 __fgets_chk F GLIBC_2.4 __fgets_unlocked_chk F diff --git a/sysdeps/unix/sysv/linux/x86_64/x32/libc.abilist b/sysdeps/unix/sysv/linux/x86_64/x32/libc.abilist index 59da85a5d8..6ff082ee4e 100644 --- a/sysdeps/unix/sysv/linux/x86_64/x32/libc.abilist +++ b/sysdeps/unix/sysv/linux/x86_64/x32/libc.abilist @@ -2158,3 +2158,4 @@ GLIBC_2.30 getdents64 F GLIBC_2.30 gettid F GLIBC_2.30 tgkill F GLIBC_2.30 twalk_r F +GLIBC_2.32 __rseq_abi T 0x20