Le 22/07/2020 à 13:10, Filip Bozuta a écrit : > Functions 'target_to_host_itimerspec()' and 'host_to_target_itimerspec()' > are used to convert values of type 'struct itimerspec' between target and > host. This type has 'struct timespec' as its fields. That is the reason > why this patch introduces a little modification to the converting functions > to be implemented using already existing functions that convert 'struct > timespec': > 'target_to_host_timespec()' and 'host_to_target_timespec()'. This makes the > code of 'target_to_host_itimerspec()' and 'host_to_target_itimerspec()' more > clean and readable. > > Signed-off-by: Filip Bozuta <[email protected]> > --- > linux-user/syscall.c | 38 ++++++++++++-------------------------- > 1 file changed, 12 insertions(+), 26 deletions(-) > > diff --git a/linux-user/syscall.c b/linux-user/syscall.c > index 1211e759c2..17e09c3cd3 100644 > --- a/linux-user/syscall.c > +++ b/linux-user/syscall.c > @@ -1236,7 +1236,9 @@ static inline abi_long copy_to_user_timeval64(abi_ulong > target_tv_addr, > defined(TARGET_NR_nanosleep) || defined(TARGET_NR_clock_settime) || \ > defined(TARGET_NR_utimensat) || defined(TARGET_NR_mq_timedsend) || \ > defined(TARGET_NR_mq_timedreceive) || defined(TARGET_NR_ipc) || \ > - defined(TARGET_NR_semop) || defined(TARGET_NR_semtimedop) > + defined(TARGET_NR_semop) || defined(TARGET_NR_semtimedop) || \ > + defined(TARGET_NR_timer_settime) || \ > + (defined(TARGET_NR_timerfd_settime) && defined(CONFIG_TIMERFD)) > static inline abi_long target_to_host_timespec(struct timespec *host_ts, > abi_ulong target_addr) > { > @@ -6790,46 +6792,30 @@ static inline abi_long target_ftruncate64(void > *cpu_env, abi_long arg1, > > #if defined(TARGET_NR_timer_settime) || \ > (defined(TARGET_NR_timerfd_settime) && defined(CONFIG_TIMERFD)) > -static inline abi_long target_to_host_itimerspec(struct itimerspec > *host_itspec, > +static inline abi_long target_to_host_itimerspec(struct itimerspec *host_its, > abi_ulong target_addr) > { > - struct target_itimerspec *target_itspec; > - > - if (!lock_user_struct(VERIFY_READ, target_itspec, target_addr, 1)) { > + if (target_to_host_timespec(&host_its->it_interval, target_addr) || > + target_to_host_timespec(&host_its->it_value, target_addr + > + sizeof(struct target_timespec))) {
The "+ sizeof(struct target_timespec)" ignores possible type alignment in the structure. I think it is safer to use "+ offsetof(struct target_timespec, it_interval)" for the first one and "+ offsetof(struct target_timespec, it_value)" for the second one. Thanks, Laurent
