From: "Tõnis Tiigi" <tonistiigi@gmail.com>
To: Laurent Vivier <laurent@vivier.eu>
Cc: qemu-devel@nongnu.org
Subject: Re: [PATCH v3 1/2] linux-user: add sched_getattr support
Date: Mon, 3 Jan 2022 09:07:56 -0800 [thread overview]
Message-ID: <CABn_tOn3aqXwPx2q173BEOTUoBjTtDLYuF34kf4mHbeAG+6+NQ@mail.gmail.com> (raw)
In-Reply-To: <CABn_tOmPjctnJU465R1b4ykRvryidj56rGeMKnaLJ+H3WAt8aw@mail.gmail.com>
Ping Laurent. Any suggestions for the follow-up questions?
On Thu, Dec 23, 2021 at 3:00 PM Tõnis Tiigi <tonistiigi@gmail.com> wrote:
>
> On Thu, Dec 23, 2021 at 1:03 PM Laurent Vivier <laurent@vivier.eu> wrote:
> >
> > Le 23/12/2021 à 07:47, Tonis Tiigi a écrit :
> >
> > Please copy here what you explain in PATCH 0 regarding this patch.
> > (do the same for PATCH 1)
> >
> > > Signed-off-by: Tonis Tiigi <tonistiigi@gmail.com>
> > > ---
> > > linux-user/syscall.c | 94 +++++++++++++++++++++++++++++++++++++++
> > > linux-user/syscall_defs.h | 14 ++++++
> > > 2 files changed, 108 insertions(+)
> > >
> > > diff --git a/linux-user/syscall.c b/linux-user/syscall.c
> > > index f1cfcc8104..2f5a0fac5a 100644
> > > --- a/linux-user/syscall.c
> > > +++ b/linux-user/syscall.c
> > > @@ -339,6 +339,12 @@ _syscall3(int, sys_sched_getaffinity, pid_t, pid, unsigned int, len,
> > > #define __NR_sys_sched_setaffinity __NR_sched_setaffinity
> > > _syscall3(int, sys_sched_setaffinity, pid_t, pid, unsigned int, len,
> > > unsigned long *, user_mask_ptr);
> > > +#define __NR_sys_sched_getattr __NR_sched_getattr
> > > +_syscall4(int, sys_sched_getattr, pid_t, pid, struct target_sched_attr *, attr,
> > > + unsigned int, size, unsigned int, flags);
> > > +#define __NR_sys_sched_setattr __NR_sched_setattr
> > > +_syscall3(int, sys_sched_setattr, pid_t, pid, struct target_sched_attr *, attr,
> > > + unsigned int, flags);
> > > #define __NR_sys_getcpu __NR_getcpu
> > > _syscall3(int, sys_getcpu, unsigned *, cpu, unsigned *, node, void *, tcache);
> > > _syscall4(int, reboot, int, magic1, int, magic2, unsigned int, cmd,
> > > @@ -10593,6 +10599,94 @@ static abi_long do_syscall1(void *cpu_env, int num, abi_long arg1,
> > > }
> > > case TARGET_NR_sched_getscheduler:
> > > return get_errno(sched_getscheduler(arg1));
> > > + case TARGET_NR_sched_getattr:
> > > + {
> > > + struct target_sched_attr *target_scha;
> > > + struct target_sched_attr scha;
> >
> > In fact, this scha is used with the host syscall, so it must be sched_attr.
>
>
> Where do you want me to define the "host variant" of sched_attr and
> with what types for the properties? Or do you want additional
> typedef(where?) so the name is less confusing? All properties in this
> type are fixed length and identical for all architectures.
>
> >
> >
> > > + if (arg2 == 0) {
> > > + return -TARGET_EINVAL;
> > > + }
> > > + if (arg3 > sizeof(scha)) {
> > > + arg3 = sizeof(scha);
> > > + }
> > > + ret = get_errno(sys_sched_getattr(arg1, &scha, arg3, arg4));
> > > + if (!is_error(ret)) {
> > > + target_scha = lock_user(VERIFY_WRITE, arg2, arg3, 0);
> > > + if (!target_scha) {
> > > + return -TARGET_EFAULT;
> > > + }
> > > + target_scha->size = tswap32(scha.size);
> > > + target_scha->sched_policy = tswap32(scha.sched_policy);
> > > + target_scha->sched_flags = tswap64(scha.sched_flags);
> > > + target_scha->sched_nice = tswap32(scha.sched_nice);
> > > + target_scha->sched_priority = tswap32(scha.sched_priority);
> > > + target_scha->sched_runtime = tswap64(scha.sched_runtime);
> > > + target_scha->sched_deadline = tswap64(scha.sched_deadline);
> > > + target_scha->sched_period = tswap64(scha.sched_period);
> > > + if (scha.size > offsetof(struct target_sched_attr, sched_util_min)) {
> > > + target_scha->sched_util_min = tswap32(scha.sched_util_min);
> > > + target_scha->sched_util_max = tswap32(scha.sched_util_max);
> > > + }
> > > + unlock_user(target_scha, arg2, arg3);
> > > + }
> > > + return ret;
> > > + }
> > > + case TARGET_NR_sched_setattr:
> > > + {
> > > + struct target_sched_attr *target_scha;
> > > + struct target_sched_attr scha;
> >
> > scha is sched_attr as it is used with the host syscall.
> >
> >
> > > + if (arg2 == 0) {
> > > + return -TARGET_EINVAL;
> > > + }
> > > + uint32_t size;
> >
> > QEMU coding style doesn't allow to mix declarations and statements.
> >
> > > + if (get_user_u32(size, arg2)) {
> > > + return -TARGET_EFAULT;
> > > + }
> > > + if (!size) {
> > > + size = offsetof(struct target_sched_attr, sched_util_min);
> > > + }
> > > + if (size < offsetof(struct target_sched_attr, sched_util_min)) {
> > > + if (put_user_u32(sizeof(struct target_sched_attr), arg2)) {
> > > + return -TARGET_EFAULT;
> > > + }
> > > + return -TARGET_E2BIG;
> > > + }
> > > +
> > > + if (size > sizeof(scha)) {
> > > + for (int i = sizeof(scha); i < size; i++) {
> > > + uint8_t b;
> > > + if (get_user_u8(b, arg2 + i)) {
> > > + return -TARGET_EFAULT;
> > > + }
> > > + if (b != 0) {
> > > + if (put_user_u32(sizeof(struct target_sched_attr), arg2)) {
> > > + return -TARGET_EFAULT;
> > > + }
> > > + return -TARGET_E2BIG;
> > > + }
> > > + }
> > > + size = sizeof(scha);
> > > + }
> >
> > I guess this is the code to mimic kernel copy_struct_from_user(), the part when usize > ksize.
> >
> > It's a little bit ugly, but I can't disagree because the kernel does the same.
> >
> > except that the kernel check for unsigned rather than for 8bit. Could you change that?
>
>
> You mean "unsigned long" like in
> https://github.com/torvalds/linux/blob/76657eaef4a759e695eb1883d4f1d9af1e4ff9a8/lib/usercopy.c#L57
> ? That would mean that this code needs to be much more complicated to
> handle the cases for the unaligned start and end bytes, need
> aligned_byte_mask helper etc. Even though kernel seems to have extra
> code for these cases iiuc it can still get EFAULT on specific
> conditions.
>
> >
> >
> > The best would be to define check_zeroed_user() in Qemu and use it here.
> >
> > > +
> > > + target_scha = lock_user(VERIFY_READ, arg2, size, 1);
> > > + if (!target_scha) {
> > > + return -TARGET_EFAULT;
> > > + }
> > > + scha.size = size;
> > > + scha.sched_policy = tswap32(target_scha->sched_policy);
> > > + scha.sched_flags = tswap64(target_scha->sched_flags);
> > > + scha.sched_nice = tswap32(target_scha->sched_nice);
> > > + scha.sched_priority = tswap32(target_scha->sched_priority);
> > > + scha.sched_runtime = tswap64(target_scha->sched_runtime);
> > > + scha.sched_deadline = tswap64(target_scha->sched_deadline);
> > > + scha.sched_period = tswap64(target_scha->sched_period);
> > > + if (size > offsetof(struct target_sched_attr, sched_util_min)) {
> > > + scha.sched_util_min = tswap32(target_scha->sched_util_min);
> > > + scha.sched_util_max = tswap32(target_scha->sched_util_max);
> > > + }
> > > + unlock_user(target_scha, arg2, 0);
> > > + return get_errno(sys_sched_setattr(arg1, &scha, arg3));
> > > + }
> > > case TARGET_NR_sched_yield:
> > > return get_errno(sched_yield());
> > > case TARGET_NR_sched_get_priority_max:
> > > diff --git a/linux-user/syscall_defs.h b/linux-user/syscall_defs.h
> > > index 0b13975937..310d6ce8ad 100644
> > > --- a/linux-user/syscall_defs.h
> > > +++ b/linux-user/syscall_defs.h
> > > @@ -2914,4 +2914,18 @@ struct target_statx {
> > > /* 0x100 */
> > > };
> > >
> > > +/* from kernel's include/linux/sched/types.h */
> > > +struct target_sched_attr {
> > > + abi_uint size;
> > > + abi_uint sched_policy;
> > > + abi_ullong sched_flags;
> > > + abi_int sched_nice;
> > > + abi_uint sched_priority;
> > > + abi_ullong sched_runtime;
> > > + abi_ullong sched_deadline;
> > > + abi_ullong sched_period;
> > > + abi_uint sched_util_min;
> > > + abi_uint sched_util_max;
> > > +};
> > > +
> > > #endif
> >
> > Thanks,
> > Laurent
next prev parent reply other threads:[~2022-01-03 17:11 UTC|newest]
Thread overview: 11+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-12-23 6:47 [PATCH v3 0/2] linux-user: fixes for sched_ syscalls Tonis Tiigi
2021-12-23 6:47 ` [PATCH v3 1/2] linux-user: add sched_getattr support Tonis Tiigi
2021-12-23 21:03 ` Laurent Vivier
2021-12-23 23:00 ` Tõnis Tiigi
2022-01-03 17:07 ` Tõnis Tiigi [this message]
2022-01-03 18:37 ` Laurent Vivier
2022-01-03 19:31 ` Tõnis Tiigi
2022-01-03 22:17 ` Laurent Vivier
2021-12-23 6:47 ` [PATCH v3 2/2] linux-user: call set/getscheduler set/getparam directly Tonis Tiigi
2021-12-23 17:44 ` Laurent Vivier
2021-12-23 20:37 ` Laurent Vivier
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=CABn_tOn3aqXwPx2q173BEOTUoBjTtDLYuF34kf4mHbeAG+6+NQ@mail.gmail.com \
--to=tonistiigi@gmail.com \
--cc=laurent@vivier.eu \
--cc=qemu-devel@nongnu.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).