lkml.org 
[lkml]   [2018]   [Aug]   [21]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [RFC][PATCH] lib/string: introduce sysfs_strncpy() and sysfs_strlcpy()
Hi Rasmus,

On (08/21/18 09:59), Rasmus Villemoes wrote:
> > +char *sysfs_strncpy(char *dest, const char *src, size_t count)
> > +{
> > + char *c;
> > +
> > + strncpy(dest, skip_spaces(src), count);
>
> I'd like to see where and how you'd use this, but I'm very skeptical of
> count being used both for the size of the dest buffer as well as an
> essentially random argument to strncpy - if count is also the maximum
> number of bytes to read from the src, you'd need to take the
> skip_spaces() into account, because there are not count bytes left after
> that...
> And if src is not necessarily nul-terminated, skip_spaces() by
> itself is wrong.

I think that sysfs input is always properly NULL-terminated. It may or
may not contain \n, but \0 is expected to be there. Am I wrong?

> Moreover, I don't think we should add more users or wrappers for strncpy
> - I highly doubt the sysfs users you have in mind want the "fill the
> rest of the buffer with '\0'" nor the "not enough room for a terminating
> '\0'? Oh well, what could possibly go wrong" semantics.

The reason I added both strncpy() and strlcpy() was that there are lots
of sysfs ->store() callbacks which use strncpy().

E.g.
channel_dimm_label_store()
dimmdev_label_store()
pmbus_add_label()
axp20x_store_attr()
cmdline_store()

and so on and on.

> > + c = dest + count - 1;
> > + while (c >= dest && (isspace(*c) || *c == '\n' || *c == '\0')) {
>
> nit: '\n' certainly already passes the isspace() test.

Hah, indeed, it should.
"\n" & 0x20 must be positive. Andrew had the same comment. But I didn't
check what actually isspace() was doing, until now.

> > +size_t sysfs_strlcpy(char *dest, const char *src, size_t size)
> > +{
> > + size_t ret;
> > + char *c;
> > +
> > + ret = strlcpy(dest, skip_spaces(src), size);
> > +
> > + size = strlen(dest);
> > + c = dest + size - 1;
> > + while (c >= dest && (isspace(*c) || *c == '\n'))
> > + c--;
> > + *(c + 1) = '\0';
> > + return ret;
> > +}
>
> What exactly is the return value?

Honestly, I didn't think about it. I wasn't sure if we want to return
anything from this function and from sysfs_strncpy(). I glanced through
a number of ->store() callbacks, and it seems that mostly people don't
bother to check strlcpy() return value at all.

> A more useful return value would either be "the length of the string
> now in dest", or some sort of indicator that the input was truncated,
> if that is ever possible.

Agreed.

> I think you're too focused on making wrappers around str[ln]cpy
> preserving parts of those functions' API. Instead, try to figure out
> what sysfs users actually want, name the functions after that, and then
> whether they use strncpy or sprintf or strscpy internally is completely
> irrelevant.

Going point, that's why the patch is in RFC stage: to figure out
what do we actually want.

> int strcpy_trim(char *dst, size_t dstsize, const char *src, size_t
> srcsize) - copy (potentially not '\0'-terminated) src to dst, trimming
> leading and trailing whitespace. dstsize must be positive, and dst is
> guaranteed to be '\0'-terminated. Returns the length of the string now
> in dst, or -EOVERFLOW if some none-whitespace character was chopped.
>
> would cover all use cases?

I like it in general. Sounds like a plan to me. Maybe the "-EOVERFLOW if
some none-whitespace character was chopped" part can be changed: if we
would trim leading and trailing whitespaces before we copy a string then
only valid input chars can get chopped.

-ss

\
 
 \ /
  Last update: 2018-08-21 11:51    [W:0.076 / U:0.076 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site