Date: Tue, 24 Mar 2015 22:37:09 +0800 From: Tiwei Bie <btw@mail.ustc.edu.cn> To: Mateusz Guzik <mjguzik@gmail.com> Cc: Konstantin Belousov <kostikbel@gmail.com>, freebsd-hackers@freebsd.org Subject: Re: [PATCH] Finish the task 'Validate coredump format string' Message-ID: <20150324143709.GA54065@freebsd> In-Reply-To: <20150324123517.GA25678@dft-labs.eu> References: <20150321200500.GC14650@dft-labs.eu> <20150322091853.GA89976@freebsd> <20150322101401.GH14650@dft-labs.eu> <20150322112555.GA44277@freebsd> <20150322113822.GB2379@kib.kiev.ua> <20150322120655.GA59757@freebsd> <20150322131524.GA95795@freebsd> <20150323005852.GB6798@dft-labs.eu> <20150323020314.GA30143@freebsd> <20150324123517.GA25678@dft-labs.eu>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, Mar 24, 2015 at 01:35:17PM +0100, Mateusz Guzik wrote:
> On Mon, Mar 23, 2015 at 10:03:14AM +0800, Tiwei Bie wrote:
> > On Mon, Mar 23, 2015 at 01:58:52AM +0100, Mateusz Guzik wrote:
> > > On Sun, Mar 22, 2015 at 09:15:24PM +0800, Tiwei Bie wrote:
> > > > On Sun, Mar 22, 2015 at 08:06:55PM +0800, Tiwei Bie wrote:
> > > > > On Sun, Mar 22, 2015 at 01:38:22PM +0200, Konstantin Belousov wrote:
> > > > > > On Sun, Mar 22, 2015 at 07:25:55PM +0800, Tiwei Bie wrote:
> > > > > > > On Sun, Mar 22, 2015 at 11:14:01AM +0100, Mateusz Guzik wrote:
> > > > > > > > On Sun, Mar 22, 2015 at 05:19:40PM +0800, Tiwei Bie wrote:
> >
[..]
> > I don't know how to do this. We need to update the man-pages?
> >
> > And sorry again, maybe I was too tired yesterday. I just find a new bug in
> > the codes above when reading your comments. Though it won't affect anything
> > cause dynamic_kenv hasn't been enabled at this time, it is really a bug.
> >
> > `format' needs to be freed, sorry... What a terrible day. ;-(
> >
>
> core(5). yeah, would be nice to mention all this stuff.
Yeah, I will update the man page!
> > I wrote these codes after consulting the codes in kern/kern_clocksource.c
> >
> [..]
> > It also needs to validate the user's inputs. And it uses a strcasecmp()
> > to check whether a different timer has been specified by the user.
> >
> > If the user is going to replace the pattern with identical one, we won't
> > need to do any update, we just `goto out'.
> >
>
> Well, I presume original author did not want to play with timers when
> changing to the same timer.
>
> Corefile is safe to change, so I suggest just dropping that strcmp.
>
> Also see below.
>
Yeah, I will remove the strcmp!
> > My new patch:
> >
> > ---
> > sys/kern/kern_sig.c | 77 ++++++++++++++++++++++++++++++++++++++++++++++++-----
> > 1 file changed, 70 insertions(+), 7 deletions(-)
> >
> > diff --git a/sys/kern/kern_sig.c b/sys/kern/kern_sig.c
> > index 8410d9d..0254517 100644
> > --- a/sys/kern/kern_sig.c
> > +++ b/sys/kern/kern_sig.c
> > @@ -3094,21 +3094,85 @@ static int compress_user_cores = 0;
> > */
> > #define corefilename_lock allproc_lock
> >
> > -static char corefilename[MAXPATHLEN] = {"%N.core"};
> > +static char corefilename[MAXPATHLEN] = "%N.core";
> > +
> > +static bool
> > +corefilename_check(const char *format)
> > +{
> > + int i, counti;
> > +
> > + counti = 0;
> > + for (i = 0; i < MAXPATHLEN && format[i] != '\0'; i++) {
> > + if (format[i] == '%') {
> > + i++;
> > + switch (format[i]) {
> > + case 'I':
> > + counti++;
> > + if (counti > 1)
> > + return (false);
>
> countI?
>
I thought `countI' seems to be CamelCase, so I named it as counti. I will
update it.
> > + case '%':
> > +}
[..]
> > +static void
> > +corefilename_init(void *arg)
> > +{
> > + char *format;
> > +
> > + format = kern_getenv("kern.corefile");
> > +
> > + if (format != NULL) {
> > + if (corefilename_check(format))
> > + strcpy(corefilename, format);
> > + freeenv(format);
> > + }
> > +}
>
> printf something. "invalid format specified" or the like.
>
> Also the code may be nicer if you
> if (format == NULL)
> return;
>
> instead
>
> > +SYSINIT(corefilename, SI_SUB_KMEM, SI_ORDER_FIRST, corefilename_init, 0);
> >
> > static int
> > sysctl_kern_corefile(SYSCTL_HANDLER_ARGS)
> > {
> > + char *format;
> > int error;
> >
> > + format = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
> > +
> > + sx_slock(&corefilename_lock);
> > + strcpy(format, corefilename);
> > + sx_sunlock(&corefilename_lock);
> > +
> > + error = sysctl_handle_string(oidp, format, MAXPATHLEN, req);
> > + if (error || !req->newptr || strcmp(format, corefilename) == 0)
> > + goto out;
> > +
>
> this should be if (error != 0 || req->newptr == NULL).
>
> strmcp was covered earlier in the mail
>
Thank you so much for your great patience! ^_^
This is my new patch:
---
share/man/man5/core.5 | 4 +++
sys/kern/kern_sig.c | 81 ++++++++++++++++++++++++++++++++++++++++++++++-----
2 files changed, 78 insertions(+), 7 deletions(-)
diff --git a/share/man/man5/core.5 b/share/man/man5/core.5
index 3f54f89..17b1ef7 100644
--- a/share/man/man5/core.5
+++ b/share/man/man5/core.5
@@ -91,6 +91,10 @@ The name defaults to
yielding the traditional
.Fx
behaviour.
+When changing the name via the
+.Va kern.corefile
+sysctl, it will fail if the new name contains
+unknown format specifiers, or its length is too long.
.Pp
By default, a process that changes user or group credentials whether
real or effective will not create a corefile.
diff --git a/sys/kern/kern_sig.c b/sys/kern/kern_sig.c
index 154c250..8cd3638 100644
--- a/sys/kern/kern_sig.c
+++ b/sys/kern/kern_sig.c
@@ -3094,21 +3094,89 @@ static int compress_user_cores = 0;
*/
#define corefilename_lock allproc_lock
-static char corefilename[MAXPATHLEN] = {"%N.core"};
+static char corefilename[MAXPATHLEN] = "%N.core";
+
+static bool
+corefilename_check(const char *format)
+{
+ int i, countI;
+
+ countI = 0;
+ for (i = 0; i < MAXPATHLEN && format[i] != '\0'; i++) {
+ if (format[i] == '%') {
+ i++;
+ switch (format[i]) {
+ case 'I':
+ countI++;
+ if (countI > 1)
+ return (false);
+ case '%':
+ case 'H':
+ case 'N':
+ case 'P':
+ case 'U':
+ break;
+ default:
+ return (false);
+ }
+ }
+ }
+
+ if (i == MAXPATHLEN)
+ return (false);
+
+ return (true);
+}
+
+static void
+corefilename_init(void *arg)
+{
+ char *format;
+
+ format = kern_getenv("kern.corefile");
+ if (format == NULL)
+ return;
+
+ if (corefilename_check(format))
+ strcpy(corefilename, format);
+ else
+ printf("Invalid format character specified in "
+ "corename `%s'\n", format);
+
+ freeenv(format);
+}
+SYSINIT(corefilename, SI_SUB_KMEM, SI_ORDER_FIRST, corefilename_init, 0);
static int
sysctl_kern_corefile(SYSCTL_HANDLER_ARGS)
{
+ char *format;
int error;
+ format = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
+
+ sx_slock(&corefilename_lock);
+ strcpy(format, corefilename);
+ sx_sunlock(&corefilename_lock);
+
+ error = sysctl_handle_string(oidp, format, MAXPATHLEN, req);
+ if (error != 0 || req->newptr == NULL)
+ goto out;
+
+ if (!corefilename_check(format)) {
+ error = EINVAL;
+ goto out;
+ }
+
sx_xlock(&corefilename_lock);
- error = sysctl_handle_string(oidp, corefilename, sizeof(corefilename),
- req);
+ strcpy(corefilename, format);
sx_xunlock(&corefilename_lock);
+out:
+ free(format, M_TEMP);
return (error);
}
-SYSCTL_PROC(_kern, OID_AUTO, corefile, CTLTYPE_STRING | CTLFLAG_RWTUN |
+SYSCTL_PROC(_kern, OID_AUTO, corefile, CTLTYPE_STRING | CTLFLAG_RW |
CTLFLAG_MPSAFE, 0, 0, sysctl_kern_corefile, "A",
"Process corefile name format string");
@@ -3171,9 +3239,8 @@ corefile_open(const char *comm, uid_t uid, pid_t pid, struct thread *td,
sbuf_printf(&sb, "%u", uid);
break;
default:
- log(LOG_ERR,
- "Unknown format character %c in "
- "corename `%s'\n", format[i], format);
+ KASSERT(0, ("Unknown format character %c in "
+ "corename `%s'\n", format[i], format));
break;
}
break;
--
2.1.2
Best regards,
Tiwei Bie
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20150324143709.GA54065>