Pintu Agarwal <pintu.ping@xxxxxxxxx> writes:
> On Thu, 6 Jan 2022 at 23:12, David Laight <David.Laight@xxxxxxxxxx> wrote:
>>
>> From: Pintu Agarwal
>> > Sent: 06 January 2022 16:50
>> >
>> > On Thu, 6 Jan 2022 at 21:41, Cyrill Gorcunov <gorcunov@xxxxxxxxx> wrote:
>> > >
>> > > On Thu, Jan 06, 2022 at 09:04:10PM +0530, Pintu Kumar wrote:
>> > > > The sysinfo member does not have any "available ram" field and
>> > > > the bufferram field is not much helpful either, to get a rough
>> > > > estimate of available ram needed for allocation.
>> > > >
>> > > > One needs to parse MemAvailable field separately from /proc/meminfo
>> > > > to get this info instead of directly getting if from sysinfo itself.
>> > > >
>> > > > Thus, this patch introduce a new field as availram in sysinfo
>> > > > so that all the info total/free/available can be retrieved from
>> > > > one place itself.
>> > > >
>> > > > There are couple of places in kernel as well where this can be improved.
>> > > > For example:
>> > > > In fs/proc/meminfo.c:
>> > > > meminfo_proc_show:
>> > > > si_meminfo(&i);
>> > > > available = si_mem_available();
>> > > > Now with this change the second call be avoided.
>> > > > Thus, we can directly do:
>> > > > show_val_kb(m, "MemAvailable: ", i.availram);
>> > > >
>> > > > Note, this also requires update in procfs for free and other commands.
>> > > > Like in free command as well we frist call sysinfo then again parse
>> > > > /proc/meminfo to get available field.
>> > > > This can be avoided too with higher kernel version.
>> > > >
>> > > > A sample output with single sysinfo call is shown below:
>> > > > Total RAM: 248376 kB
>> > > > Free RAM: 231540 kB
>> > > > Avail RAM: 230448 kB
>> > > >
>> > > > Signed-off-by: Pintu Kumar <quic_pintu@xxxxxxxxxxx>
>> > > > Signed-off-by: Pintu Agarwal <pintu.ping@xxxxxxxxx>
>> > > > ---
>> > > > include/uapi/linux/sysinfo.h | 1 +
>> > > > kernel/sys.c | 4 ++++
>> > > > mm/page_alloc.c | 2 ++
>> > > > 3 files changed, 7 insertions(+)
>> > > >
>> > > > diff --git a/include/uapi/linux/sysinfo.h b/include/uapi/linux/sysinfo.h
>> > > > index 435d5c2..6e77e90 100644
>> > > > --- a/include/uapi/linux/sysinfo.h
>> > > > +++ b/include/uapi/linux/sysinfo.h
>> > > > @@ -12,6 +12,7 @@ struct sysinfo {
>> > > > __kernel_ulong_t freeram; /* Available memory size */
>> > > > __kernel_ulong_t sharedram; /* Amount of shared memory */
>> > > > __kernel_ulong_t bufferram; /* Memory used by buffers */
>> > > > + __kernel_ulong_t availram; /* Memory available for allocation */
>> > > > __kernel_ulong_t totalswap; /* Total swap space size */
>> > > > __kernel_ulong_t freeswap; /* swap space still available */
>> > > > __u16 procs; /* Number of current processes */
>> > >
>> > > Hi! Sorry, but I don't understand -- the sysinfo structure seems to
>> > > be part of user API, no? Don't we break it up here?
>> >
>> > Yes, the corresponding user space header /usr/include/linux/sysinfo.h
>> > also needs to be updated.
>> > When we generate the kernel header it will be updated automatically.
>>
>> You can't add a field in the middle of a UAPI structure.
>> It breaks compatibility for old binaries.
>>
>> Depending on the interface definition you may be able to add one at the end.
>>
> oh okay thank you for your feedback. I will move to the end and check again.
> But my doubt is, whether I should move before this
> char _f[20-2*sizeof(__kernel_ulong_t)-sizeof(__u32)];
> or after this ?
Before the padding and you should reduce the size of the padding by the
size of your new field.
> Also, I could not understand what this is for ?
> Do we need to update this since sture is changed ?
In general padding like that is so new fields can be added. The
comment about libc5 makes me a wonder a bit, but I expect libc5 just
added the padding in it's copy of the structure that it exported to
userspace many many years ago so that new fields could be added.
Eric
