From: Petr Mladek <pmladek@suse.com> To: Greg KH <gregkh@linuxfoundation.org> Cc: kernel-hardening@lists.openwall.com, Sergey Senozhatsky <sergey.senozhatsky@gmail.com>, linux-kernel@vger.kernel.org, Catalin Marinas <catalin.marinas@arm.com>, Will Deacon <will.deacon@arm.com>, Steven Rostedt <rostedt@goodmis.org>, William Roberts <william.c.roberts@intel.com>, Chris Fries <cfries@google.com>, Dave Weinstein <olorin@google.com> Subject: Re: [RFC 1/6] lib: vsprintf: additional kernel pointer filtering options Date: Tue, 16 May 2017 13:58:11 +0200 [thread overview] Message-ID: <20170516115811.GA7297@pathway.suse.cz> (raw) In-Reply-To: <20170506040656.GB32707@kroah.com> On Fri 2017-05-05 21:06:56, Greg KH wrote: > From: Dave Weinstein <olorin@google.com> > > Add the kptr_restrict setting of 3 which results in both > %p and %pK values being replaced by zeros. > > Add an additional %pP value inspired by the Grsecurity > option which explicitly whitelists pointers for output. > > This patch is based on work by William Roberts > <william.c.roberts@intel.com> > > diff --git a/lib/vsprintf.c b/lib/vsprintf.c > index e3bf4e0f10b5..f4e11dade1ab 100644 > --- a/lib/vsprintf.c > +++ b/lib/vsprintf.c > @@ -395,6 +395,16 @@ struct printf_spec { > #define FIELD_WIDTH_MAX ((1 << 23) - 1) > #define PRECISION_MAX ((1 << 15) - 1) > > +int kptr_restrict __read_mostly; > + > +/* > + * Always cleanse %p and %pK specifiers > + */ > +static inline int kptr_restrict_always_cleanse_pointers(void) The name of the function is very long and still confusing. It uses the word "always" but there are many types of pointers that are not cleared with this condition, for example %pP, %pa. Do we need this helper function at all? It is used a weird way, see below. > +{ > + return kptr_restrict >= 3; > +} > + > static noinline_for_stack > char *number(char *buf, char *end, unsigned long long num, > struct printf_spec spec) > @@ -1576,7 +1588,7 @@ char *pointer(const char *fmt, char *buf, char *end, void *ptr, > { > const int default_width = 2 * sizeof(void *); > > - if (!ptr && *fmt != 'K') { > + if (!ptr && *fmt != 'K' && !kptr_restrict_always_cleanse_pointers()) { > /* > * Print (null) with the same width as a pointer so it makes > * tabular output look nice. > @@ -1657,10 +1669,43 @@ char *pointer(const char *fmt, char *buf, char *end, void *ptr, > va_end(va); > return buf; > } > + case 'N': > + return netdev_bits(buf, end, ptr, fmt); > + case 'a': > + return address_val(buf, end, ptr, fmt); > + case 'd': > + return dentry_name(buf, end, ptr, spec, fmt); > + case 'C': > + return clock(buf, end, ptr, spec, fmt); > + case 'D': > + return dentry_name(buf, end, > + ((const struct file *)ptr)->f_path.dentry, > + spec, fmt); > +#ifdef CONFIG_BLOCK > + case 'g': > + return bdev_name(buf, end, ptr, spec, fmt); > +#endif > + > + case 'G': > + return flags_string(buf, end, ptr, fmt); > + case 'P': > + /* > + * an explicitly whitelisted kernel pointer should never be > + * cleansed > + */ > + break; > + default: > + /* > + * plain %p, no extension, check if we should always cleanse and > + * treat like %pK. > + */ > + if (!kptr_restrict_always_cleanse_pointers()) > + break; > + /* fallthrough */ Using default: in the middle of other cases is pretty confusing and a call for troubles. > case 'K': > switch (kptr_restrict) { > case 0: > - /* Always print %pK values */ > + /* Always print %p values */ And this is one example. We are never here for %p because we called break in the "default:" case above for kptr_restrict == 0. > break; > case 1: { > const struct cred *cred; > @@ -1679,7 +1724,7 @@ char *pointer(const char *fmt, char *buf, char *end, void *ptr, > * Only print the real pointer value if the current > * process has CAP_SYSLOG and is running with the > * same credentials it started with. This is because > - * access to files is checked at open() time, but %pK > + * access to files is checked at open() time, > but %p Same here. This change makes the feeling that we check CAP_SYSLOG even for plain %p but we actually do not do it. > * checks permission at read() time. We don't want to > * leak pointer values if a binary opens a file using > * %pK and then elevates privileges before reading it. > @@ -1691,33 +1736,13 @@ char *pointer(const char *fmt, char *buf, char *end, void *ptr, > ptr = NULL; > break; > } > - case 2: > + case 2: /* restrict only %pK */ > + case 3: /* restrict all non-extensioned %p and %pK */ > default: > - /* Always print 0's for %pK */ > ptr = NULL; > break; > } > break; I would personally write this piece of code a more straightforward way, for example: switch (*fmt) { [...] case 'P': /* Always print an explicitly whitelisted kernel pointer. */ break; case 'K': /* Always print %pK values when there are no restrictions. */ if (!kptr_restrict) break; /* Cleanse %pK values for non-privileged users. */ if (kptr_restrict == 1) { const struct cred *cred; /* * kptr_restrict==1 cannot be used in IRQ context * because its test for CAP_SYSLOG would be meaningless. */ if (in_irq() || in_serving_softirq() || in_nmi()) { if (spec.field_width == -1) spec.field_width = default_width; return string(buf, end, "pK-error", spec); } /* * Only print the real pointer value if the current * process has CAP_SYSLOG and is running with the * same credentials it started with. This is because * access to files is checked at open() time, but %p * checks permission at read() time. We don't want to * leak pointer values if a binary opens a file using * %pK and then elevates privileges before reading it. */ cred = current_cred(); if (!has_capability_noaudit(current, CAP_SYSLOG) || !uid_eq(cred->euid, cred->uid) || !gid_eq(cred->egid, cred->gid)) ptr = NULL; break; } /* Always cleanse %pK values in higher restrition levels. */ ptr = NULL; break; default: /* * Also plain pointers (%p) are always cleansed in higher * restriction levels. */ if (kptr_restrict >= 3) ptr = NULL; } Best Regards, Petr
WARNING: multiple messages have this Message-ID (diff)
From: Petr Mladek <pmladek@suse.com> To: Greg KH <gregkh@linuxfoundation.org> Cc: kernel-hardening@lists.openwall.com, Sergey Senozhatsky <sergey.senozhatsky@gmail.com>, linux-kernel@vger.kernel.org, Catalin Marinas <catalin.marinas@arm.com>, Will Deacon <will.deacon@arm.com>, Steven Rostedt <rostedt@goodmis.org>, William Roberts <william.c.roberts@intel.com>, Chris Fries <cfries@google.com>, Dave Weinstein <olorin@google.com> Subject: [kernel-hardening] Re: [RFC 1/6] lib: vsprintf: additional kernel pointer filtering options Date: Tue, 16 May 2017 13:58:11 +0200 [thread overview] Message-ID: <20170516115811.GA7297@pathway.suse.cz> (raw) In-Reply-To: <20170506040656.GB32707@kroah.com> On Fri 2017-05-05 21:06:56, Greg KH wrote: > From: Dave Weinstein <olorin@google.com> > > Add the kptr_restrict setting of 3 which results in both > %p and %pK values being replaced by zeros. > > Add an additional %pP value inspired by the Grsecurity > option which explicitly whitelists pointers for output. > > This patch is based on work by William Roberts > <william.c.roberts@intel.com> > > diff --git a/lib/vsprintf.c b/lib/vsprintf.c > index e3bf4e0f10b5..f4e11dade1ab 100644 > --- a/lib/vsprintf.c > +++ b/lib/vsprintf.c > @@ -395,6 +395,16 @@ struct printf_spec { > #define FIELD_WIDTH_MAX ((1 << 23) - 1) > #define PRECISION_MAX ((1 << 15) - 1) > > +int kptr_restrict __read_mostly; > + > +/* > + * Always cleanse %p and %pK specifiers > + */ > +static inline int kptr_restrict_always_cleanse_pointers(void) The name of the function is very long and still confusing. It uses the word "always" but there are many types of pointers that are not cleared with this condition, for example %pP, %pa. Do we need this helper function at all? It is used a weird way, see below. > +{ > + return kptr_restrict >= 3; > +} > + > static noinline_for_stack > char *number(char *buf, char *end, unsigned long long num, > struct printf_spec spec) > @@ -1576,7 +1588,7 @@ char *pointer(const char *fmt, char *buf, char *end, void *ptr, > { > const int default_width = 2 * sizeof(void *); > > - if (!ptr && *fmt != 'K') { > + if (!ptr && *fmt != 'K' && !kptr_restrict_always_cleanse_pointers()) { > /* > * Print (null) with the same width as a pointer so it makes > * tabular output look nice. > @@ -1657,10 +1669,43 @@ char *pointer(const char *fmt, char *buf, char *end, void *ptr, > va_end(va); > return buf; > } > + case 'N': > + return netdev_bits(buf, end, ptr, fmt); > + case 'a': > + return address_val(buf, end, ptr, fmt); > + case 'd': > + return dentry_name(buf, end, ptr, spec, fmt); > + case 'C': > + return clock(buf, end, ptr, spec, fmt); > + case 'D': > + return dentry_name(buf, end, > + ((const struct file *)ptr)->f_path.dentry, > + spec, fmt); > +#ifdef CONFIG_BLOCK > + case 'g': > + return bdev_name(buf, end, ptr, spec, fmt); > +#endif > + > + case 'G': > + return flags_string(buf, end, ptr, fmt); > + case 'P': > + /* > + * an explicitly whitelisted kernel pointer should never be > + * cleansed > + */ > + break; > + default: > + /* > + * plain %p, no extension, check if we should always cleanse and > + * treat like %pK. > + */ > + if (!kptr_restrict_always_cleanse_pointers()) > + break; > + /* fallthrough */ Using default: in the middle of other cases is pretty confusing and a call for troubles. > case 'K': > switch (kptr_restrict) { > case 0: > - /* Always print %pK values */ > + /* Always print %p values */ And this is one example. We are never here for %p because we called break in the "default:" case above for kptr_restrict == 0. > break; > case 1: { > const struct cred *cred; > @@ -1679,7 +1724,7 @@ char *pointer(const char *fmt, char *buf, char *end, void *ptr, > * Only print the real pointer value if the current > * process has CAP_SYSLOG and is running with the > * same credentials it started with. This is because > - * access to files is checked at open() time, but %pK > + * access to files is checked at open() time, > but %p Same here. This change makes the feeling that we check CAP_SYSLOG even for plain %p but we actually do not do it. > * checks permission at read() time. We don't want to > * leak pointer values if a binary opens a file using > * %pK and then elevates privileges before reading it. > @@ -1691,33 +1736,13 @@ char *pointer(const char *fmt, char *buf, char *end, void *ptr, > ptr = NULL; > break; > } > - case 2: > + case 2: /* restrict only %pK */ > + case 3: /* restrict all non-extensioned %p and %pK */ > default: > - /* Always print 0's for %pK */ > ptr = NULL; > break; > } > break; I would personally write this piece of code a more straightforward way, for example: switch (*fmt) { [...] case 'P': /* Always print an explicitly whitelisted kernel pointer. */ break; case 'K': /* Always print %pK values when there are no restrictions. */ if (!kptr_restrict) break; /* Cleanse %pK values for non-privileged users. */ if (kptr_restrict == 1) { const struct cred *cred; /* * kptr_restrict==1 cannot be used in IRQ context * because its test for CAP_SYSLOG would be meaningless. */ if (in_irq() || in_serving_softirq() || in_nmi()) { if (spec.field_width == -1) spec.field_width = default_width; return string(buf, end, "pK-error", spec); } /* * Only print the real pointer value if the current * process has CAP_SYSLOG and is running with the * same credentials it started with. This is because * access to files is checked at open() time, but %p * checks permission at read() time. We don't want to * leak pointer values if a binary opens a file using * %pK and then elevates privileges before reading it. */ cred = current_cred(); if (!has_capability_noaudit(current, CAP_SYSLOG) || !uid_eq(cred->euid, cred->uid) || !gid_eq(cred->egid, cred->gid)) ptr = NULL; break; } /* Always cleanse %pK values in higher restrition levels. */ ptr = NULL; break; default: /* * Also plain pointers (%p) are always cleansed in higher * restriction levels. */ if (kptr_restrict >= 3) ptr = NULL; } Best Regards, Petr
next prev parent reply other threads:[~2017-05-16 11:58 UTC|newest] Thread overview: 35+ messages / expand[flat|nested] mbox.gz Atom feed top 2017-05-06 4:06 [RFC 00/06] printk: add more new kernel pointer filter options Greg KH 2017-05-06 4:06 ` [kernel-hardening] " Greg KH 2017-05-06 4:06 ` [RFC 1/6] lib: vsprintf: additional kernel pointer filtering options Greg KH 2017-05-06 4:06 ` [kernel-hardening] " Greg KH 2017-05-16 11:58 ` Petr Mladek [this message] 2017-05-16 11:58 ` [kernel-hardening] " Petr Mladek 2017-05-18 14:12 ` Greg KH 2017-05-18 14:12 ` [kernel-hardening] " Greg KH 2017-05-06 4:07 ` [RFC 2/6] lib: vsprintf: whitelist stack traces Greg KH 2017-05-06 4:07 ` [kernel-hardening] " Greg KH 2017-05-06 4:07 ` [RFC 3/6] lib: vsprintf: physical address kernel pointer filtering options Greg KH 2017-05-06 4:07 ` [kernel-hardening] " Greg KH 2017-05-06 10:48 ` Ian Campbell 2017-05-06 4:07 ` [RFC 4/6] lib: vsprintf: default kptr_restrict to the maximum value Greg KH 2017-05-06 4:07 ` [kernel-hardening] " Greg KH 2017-05-06 4:07 ` [RFC 5/6] lib: vsprintf: Add "%paP", "%padP" options Greg KH 2017-05-06 4:07 ` [kernel-hardening] " Greg KH 2017-05-06 4:42 ` Joe Perches 2017-05-06 4:42 ` [kernel-hardening] " Joe Perches 2017-05-06 5:00 ` Greg KH 2017-05-06 5:00 ` [kernel-hardening] " Greg KH 2017-05-16 14:41 ` Petr Mladek 2017-05-16 14:41 ` [kernel-hardening] " Petr Mladek 2017-05-18 14:12 ` Greg KH 2017-05-18 14:12 ` [kernel-hardening] " Greg KH 2017-05-06 4:07 ` [RFC 6/6] drivers: uio: Un-restrict sysfs pointers for UIO Greg KH 2017-05-06 4:07 ` [kernel-hardening] " Greg KH 2017-05-11 1:37 ` [RFC 00/06] printk: add more new kernel pointer filter options Sergey Senozhatsky 2017-05-11 1:37 ` [kernel-hardening] " Sergey Senozhatsky 2017-05-16 21:36 ` Roberts, William C 2017-05-16 21:36 ` [kernel-hardening] " Roberts, William C 2017-05-18 14:13 ` Greg KH 2017-05-18 14:13 ` [kernel-hardening] " Greg KH 2017-05-19 20:25 ` Roberts, William C 2017-05-19 20:25 ` [kernel-hardening] " Roberts, William C
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=20170516115811.GA7297@pathway.suse.cz \ --to=pmladek@suse.com \ --cc=catalin.marinas@arm.com \ --cc=cfries@google.com \ --cc=gregkh@linuxfoundation.org \ --cc=kernel-hardening@lists.openwall.com \ --cc=linux-kernel@vger.kernel.org \ --cc=olorin@google.com \ --cc=rostedt@goodmis.org \ --cc=sergey.senozhatsky@gmail.com \ --cc=will.deacon@arm.com \ --cc=william.c.roberts@intel.com \ /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: linkBe sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.