add support for never printing hashed addresses
Subsystems (5)
cover add support for never printing hashed addresses
[accidentally sent from the wrong email address, so resending]
[The list of email addresses on CC: is getting quite lengthy, so I hope I've included everyone.]
Although hashing addresses printed via printk does make the kernel more secure, it interferes with debugging, especially with some functions like print_hex_dump() which always uses hashed addresses.
To avoid having to choose between %p and %px, it's easier to add a kernel command line that treats all %p as %px. This encourages developers to use %p more without making debugging more difficult.
Patches #1 and #2 upgrade the kselftest framework so that it can report on tests that were skipped outright. This is needed for the test_printf module which will now skip %p hashing tests if hashing is disabled.
Patch #2 upgrades the printf library to check the command line. It also updates test_printf().
Timur Tabi (3):
lib/test_printf: use KSTM_MODULE_GLOBALS macro
kselftest: add support for skipped tests
[v2] lib/vsprintf: make-printk-non-secret printks all addresses as
unhashed
.../admin-guide/kernel-parameters.txt | 15 +++++++ lib/test_printf.c | 12 +++++- lib/vsprintf.c | 40 ++++++++++++++++++- tools/testing/selftests/kselftest_module.h | 18 ++++++--- 4 files changed, 75 insertions(+), 10 deletions(-)
Signature · 1 line
2.25.1
PATCH 1/3 lib/test_printf: use KSTM_MODULE_GLOBALS macro
Instead of defining the total/failed test counters manually, test_printf should use the kselftest macro created for this purpose.
--- lib/test_printf.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
lib/test_printf.c +2 -2
@@ -30,8 +30,8 @@#define PAD_SIZE 16#define FILL_CHAR '$' -static unsigned total_tests__initdata ; -static unsigned failed_tests__initdata ; +KSTM_MODULE_GLOBALS (); +static char *test_buffer__initdata ;static char *alloced_buffer__initdata ;
Signature · 1 line
2.25.1
Re: [PATCH 1/3] lib/test_printf: use KSTM_MODULE_GLOBALS macro
Quoted 5 lines · Timur Tabi
Instead of defining the total/failed test counters manually,
test_printfshould use the kselftest macro created for this purpose.Signed-off-by Timur Tabi<[email protected]>
Ugh, I really need to stop sending patches late at night. This is again the wrong email address.
I'm sure I'll need to post another version of these patches, so I'll just fix it then.
Re: [PATCH 1/3] lib/test_printf: use KSTM_MODULE_GLOBALS macro
In reply to Timur TabiQuoted 5 lines · Timur Tabi
Instead of defining the total/failed test counters manually,
test_printfshould use the kselftest macro created for this purpose.Signed-off-by Timur Tabi <[email protected]>
Best Regards,
Petr
PATCH 2/3 kselftest: add support for skipped tests
In reply to Timur TabiUpdate the kselftest framework to all testing clients to specify that some tests were skipped.
--- tools/testing/selftests/kselftest_module.h | 18 ++++++++++++------ 1 file changed, 12 insertions(+), 6 deletions(-)
tools/testing/selftests/kselftest_module.h +12 -6
@@ -11,7 +11,8 @@#define KSTM_MODULE_GLOBALS () \ static unsigned int total_tests __initdata; \ -static unsigned int failed_tests __initdata +static unsigned int failed_tests __initdata; \ +static unsigned int skipped_tests __initdata#define KSTM_CHECK_ZERO (x) do { \ total_tests++; \
@@ -21,11 +22,16 @@ static unsigned int failed_tests __initdata } \ }while (0 ) -static inline int kstm_report(unsigned int total_tests, unsigned int failed_tests) +static inline int kstm_report(unsigned int total_tests, unsigned int failed_tests, +unsigned int skipped_tests) { -if (failed_tests ==0 ) -pr_info ("all %u tests passed\n" , total_tests); -else +if (failed_tests ==0 ) { +if (skipped_tests) { +pr_info ("skipped %u tests\n" , skipped_tests); +pr_info ("remaining %u tests passed\n" , total_tests); + }else +pr_info ("all %u tests passed\n" , total_tests); + }else pr_warn ("failed %u out of %u tests\n" , failed_tests, total_tests);return failed_tests ? -EINVAL :0 ;
@@ -36,7 +42,7 @@ static int __init __module##_init(void) \ { \pr_info ("loaded.\n" ); \selftest (); \ -return kstm_report (total_tests, failed_tests); \ +return kstm_report (total_tests, failed_tests, skipped_tests); \ } \ static void __exit __module##_exit(void) \ { \
Signature · 1 line
2.25.1
kernel test robot
build botRe: [PATCH 2/3] kselftest: add support for skipped tests
I love your patch! Yet something to improve:
Hi Timur,
I love your patch! Yet something to improve:
[auto build test ERROR on kselftest/next]
[also build test ERROR on linus/master hnaz-linux-mm/master v5.11-rc7 next-20210125] [If your patch is applied to the wrong git tree, kindly drop us a note. And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch]
url: https://github.com/0day-ci/linux/commits/Timur-Tabi/add-support-for-never-printing-hashed-addresses/20210210-131927 base: https://git.kernel.org/pub/scm/linux/kernel/git/shuah/linux-kselftest.git next config: h8300-randconfig-s031-20210209 (attached as .config) compiler: h8300-linux-gcc (GCC) 9.3.0
reproduce:
wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross chmod +x ~/bin/make.cross
# apt-get install sparse
# sparse version: v0.6.3-215-g0fb77bb6-dirty
# https://github.com/0day-ci/linux/commit/4387344bd9c51be401880f17c193e4956036c067
git remote add linux-review https://github.com/0day-ci/linux
git fetch --no-tags linux-review Timur-Tabi/add-support-for-never-printing-hashed-addresses/20210210-131927 git checkout 4387344bd9c51be401880f17c193e4956036c067
# save the attached .config to linux build tree
COMPILER_INSTALL_PATH=$HOME/0day COMPILER=gcc-9.3.0 make.cross C=1 CF='-fdiagnostic-prefix -D__CHECK_ENDIAN__' ARCH=h8300
If you fix the issue, kindly add following tag as appropriate
All errors (new ones prefixed by >>):
In file included from lib/test_bitmap.c:17:
lib/test_bitmap.c: In function 'test_bitmap_init':
>> lib/../tools/testing/selftests/kselftest_module.h:45:48: error: 'skipped_tests' undeclared (first use in this function); did you mean 'failed_tests'?
45 | return kstm_report(total_tests, failed_tests, skipped_tests); \
| ^~~~~~~~~~~~~
lib/test_bitmap.c:637:1: note: in expansion of macro 'KSTM_MODULE_LOADERS'
637 | KSTM_MODULE_LOADERS(test_bitmap);
| ^~~~~~~~~~~~~~~~~~~
lib/../tools/testing/selftests/kselftest_module.h:45:48: note: each undeclared identifier is reported only once for each function it appears in
45 | return kstm_report(total_tests, failed_tests, skipped_tests); \
| ^~~~~~~~~~~~~
lib/test_bitmap.c:637:1: note: in expansion of macro 'KSTM_MODULE_LOADERS'
637 | KSTM_MODULE_LOADERS(test_bitmap);
| ^~~~~~~~~~~~~~~~~~~
lib/test_bitmap.c:637:1: error: control reaches end of non-void function [-Werror=return-type]
637 | KSTM_MODULE_LOADERS(test_bitmap);
| ^~~~~~~~~~~~~~~~~~~cc1: some warnings being treated as errors
vim +45 lib/../tools/testing/selftests/kselftest_module.h
39
40 #define KSTM_MODULE_LOADERS(__module) \
41 static int __init __module##_init(void) \
42 { \
43 pr_info("loaded.\n"); \
44 selftest(); \
> 45 return kstm_report(total_tests, failed_tests, skipped_tests); \
46 } \
47 static void __exit __module##_exit(void) \
48 { \
49 pr_info("unloaded.\n"); \
50 } \
51 module_init(__module##_init); \
52 module_exit(__module##_exit)
53 ---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all(a)lists.01.org
PATCH 3/3 lib/vsprintf: make-printk-non-secret printks all addresses as unhashed
In reply to Timur TabiIf the make-printk-non-secret command line parameter is set, then printk("%p") will print pointers as unhashed. This is useful for debugging purposes.
A large warning message is displayed if this option is enabled. Unhashed pointers, while useful for debugging, expose kernel addresses which can be a security risk.
Also update test_printf to skip the hashed pointer tests if the command-line option is set.
--- .../admin-guide/kernel-parameters.txt | 15 ++++++++ lib/test_printf.c | 8 ++++ lib/vsprintf.c | 38 ++++++++++++++++++- 3 files changed, 59 insertions(+), 2 deletions(-)
Documentation/admin-guide/kernel-parameters.txt +15 -0
@@ -2613,6 +2613,21 @@ different yeeloong laptops. Example: machtype=lemote-yeeloong-2f-7inch + make-printk-non-secret + Force pointers printed to the console to be unhashed. + By default, when a pointer is printed to the kernel + console (via %p format string), that pointer is + "hashed", i.e. obscured by hashing the pointer value. + This is a security feature that hides actual kernel + addresses from unprivileged users, but it also makes + debugging the kernel more difficult since unequal + pointers can no longer be compared. If this option is + specified, then all normal pointers will have their + true value printed. Pointers printed via %pK may + still be hashed. This option should only be specified + when debugging the kernel. Please do not use on + production kernels. + max_addr=nn[KMG] [KNL,BOOT,ia64] All physical memory greater than or equal to this physical address is ignored.
lib/test_printf.c +8 -0
@@ -35,6 +35,8 @@ KSTM_MODULE_GLOBALS();static char *test_buffer__initdata ;static char *alloced_buffer__initdata ; +extern bool debug_never_hash_pointers; + static int __printf(4, 0) __init do_test(int bufsize, const char *expect, int elen, const char *fmt, va_list ap)
@@ -301,6 +303,12 @@ plain(void) {int err; +if (debug_never_hash_pointers) { +pr_warn ("skipping plain 'p' tests" ); + skipped_tests +=2 ; +return ; + } + err =plain_hash ();if (err) {pr_warn ("plain 'p' does not appear to be hashed\n" );
lib/vsprintf.c +36 -2
@@ -2090,6 +2090,34 @@ char *fwnode_string(char *buf, char *end, struct fwnode_handle *fwnode,return widen_string (buf, buf - buf_start, end, spec); } +/* Disable pointer hashing if requested */ +bool debug_never_hash_pointers__ro_after_init ; +EXPORT_SYMBOL_GPL (debug_never_hash_pointers); + +static int __init debug_never_hash_pointers_enable (char *str) +{ + debug_never_hash_pointers = true; + +pr_warn ("**********************************************************\n" ); +pr_warn ("** NOTICE NOTICE NOTICE NOTICE NOTICE NOTICE NOTICE **\n" ); +pr_warn ("** **\n" ); +pr_warn ("** All pointers that are printed to the console will **\n" ); +pr_warn ("** be printed as unhashed. **\n" ); +pr_warn ("** **\n" ); +pr_warn ("** Kernel memory addresses are exposed, which may **\n" ); +pr_warn ("** reduce the security of your system. **\n" ); +pr_warn ("** **\n" ); +pr_warn ("** If you see this message and you are not debugging **\n" ); +pr_warn ("** the kernel, report this immediately to your system **\n" ); +pr_warn ("** administrator! **\n" ); +pr_warn ("** **\n" ); +pr_warn ("** NOTICE NOTICE NOTICE NOTICE NOTICE NOTICE NOTICE **\n" ); +pr_warn ("**********************************************************\n" ); + +return 0 ; +} +early_param ("make-printk-non-secret" , debug_never_hash_pointers_enable); +/* * Show a '%p' thing. A kernel extension is that the '%p' is followed * by an extra set of alphanumeric characters that are extended format
@@ -2297,8 +2325,14 @@ char *pointer(const char *fmt, char *buf, char *end, void *ptr, } } -/* default is to _not_ leak addresses, hash before printing */ -return ptr_to_id (buf, end, ptr, spec); +/* +* default is to _not_ leak addresses, so hash before printing unless +* make-printk-non-secret is specified on the command line. +*/ +if (unlikely (debug_never_hash_pointers)) +return pointer_string (buf, end, ptr, spec); +else +return ptr_to_id (buf, end, ptr, spec); }/*
Signature · 1 line
2.25.1
Re: [PATCH 3/3] lib/vsprintf: make-printk-non-secret printks all addresses as unhashed
Quoted 8 lines · Timur Tabi
If the make-printk-non-secret command line parameter is set, then
printk("%p")will print pointers as unhashed. This is useful for debugging purposes.A large warning message is displayed if this option is enabled. Unhashed pointers, while useful for debugging, expose kernel addresses which can be a security risk.
Quoted 7 lines
Also update
test_printfto skip the hashed pointer tests if the command-line option is set.Signed-off-by Timur Tabi <[email protected]>Acked-by Petr Mladek <[email protected]>Acked-by Randy Dunlap <[email protected]>
Thanks!
Quoted 115 lines
--- .../admin-guide/kernel-parameters.txt | 15 ++++++++ lib/test_printf.c | 8 ++++ lib/vsprintf.c | 38 ++++++++++++++++++- 3 files changed, 59 insertions(+), 2 deletions(-)Documentation/admin-guide/kernel-parameters.txt +15
-0@@ -2613,6 +2613,21 @@ different yeeloong laptops. Example: machtype=lemote-yeeloong-2f-7inch + make-printk-non-secret + Force pointers printed to the console to be unhashed. + By default, when a pointer is printed to the kernel + console (via %p format string), that pointer is + "hashed", i.e. obscured by hashing the pointer value. + This is a security feature that hides actual kernel + addresses from unprivileged users, but it also makes + debugging the kernel more difficult since unequal + pointers can no longer be compared. If this option is + specified, then all normal pointers will have their + true value printed. Pointers printed via %pK may + still be hashed. This option should only be specified + when debugging the kernel. Please do not use on + production kernels. + max_addr=nn[KMG] [KNL,BOOT,ia64] All physical memory greater than or equal to this physical address is ignored.lib/test_printf.c +8
-0@@ -35,6 +35,8 @@ KSTM_MODULE_GLOBALS();static char *test_buffer__initdata ;static char *alloced_buffer__initdata ; +extern bool debug_never_hash_pointers; + static int __printf(4, 0) __init do_test(int bufsize, const char *expect, int elen, const char *fmt, va_list ap)@@ -301,6 +303,12 @@ plain(void) {int err; +if (debug_never_hash_pointers) { +pr_warn ("skipping plain 'p' tests" ); + skipped_tests +=2 ; +return ; + } + err =plain_hash ();if (err) {pr_warn ("plain 'p' does not appear to be hashed\n" );lib/vsprintf.c +36
-2@@ -2090,6 +2090,34 @@ char *fwnode_string(char *buf, char *end, struct fwnode_handle *fwnode,return widen_string (buf, buf - buf_start, end, spec); } +/* Disable pointer hashing if requested */ +bool debug_never_hash_pointers__ro_after_init ; +EXPORT_SYMBOL_GPL (debug_never_hash_pointers); + +static int __init debug_never_hash_pointers_enable (char *str) +{ + debug_never_hash_pointers = true; + +pr_warn ("**********************************************************\n" ); +pr_warn ("** NOTICE NOTICE NOTICE NOTICE NOTICE NOTICE NOTICE **\n" ); +pr_warn ("** **\n" ); +pr_warn ("** All pointers that are printed to the console will **\n" ); +pr_warn ("** be printed as unhashed. **\n" ); +pr_warn ("** **\n" ); +pr_warn ("** Kernel memory addresses are exposed, which may **\n" ); +pr_warn ("** reduce the security of your system. **\n" ); +pr_warn ("** **\n" ); +pr_warn ("** If you see this message and you are not debugging **\n" ); +pr_warn ("** the kernel, report this immediately to your system **\n" ); +pr_warn ("** administrator! **\n" ); +pr_warn ("** **\n" ); +pr_warn ("** NOTICE NOTICE NOTICE NOTICE NOTICE NOTICE NOTICE **\n" ); +pr_warn ("**********************************************************\n" ); + +return 0 ; +} +early_param ("make-printk-non-secret" , debug_never_hash_pointers_enable); +/* * Show a '%p' thing. A kernel extension is that the '%p' is followed * by an extra set of alphanumeric characters that are extended format @@ -2297,8 +2325,14 @@ char *pointer(const char *fmt, char *buf, char *end, void *ptr, } } -/* default is to _not_ leak addresses, hash before printing */ -return ptr_to_id (buf, end, ptr, spec); +/* +* default is to _not_ leak addresses, so hash before printing unless +* make-printk-non-secret is specified on the command line. +*/ +if (unlikely (debug_never_hash_pointers)) +return pointer_string (buf, end, ptr, spec); +else +return ptr_to_id (buf, end, ptr, spec); }/*
Petr Mladek
maintainerRe: [PATCH 3/3] lib/vsprintf: make-printk-non-secret printks all addresses as unhashed
In reply to Timur TabiQuoted 98 lines · Timur Tabi
If the make-printk-non-secret command line parameter is set, then
printk("%p")will print pointers as unhashed. This is useful for debugging purposes.A large warning message is displayed if this option is enabled. Unhashed pointers, while useful for debugging, expose kernel addresses which can be a security risk.
Also update
test_printfto skip the hashed pointer tests if the command-line option is set.Signed-off-by Timur Tabi <[email protected]>Acked-by Petr Mladek <[email protected]>Acked-by Randy Dunlap <[email protected]>--- .../admin-guide/kernel-parameters.txt | 15 ++++++++ lib/test_printf.c | 8 ++++ lib/vsprintf.c | 38 ++++++++++++++++++- 3 files changed, 59 insertions(+), 2 deletions(-)Documentation/admin-guide/kernel-parameters.txt +15
-0@@ -2613,6 +2613,21 @@ different yeeloong laptops. Example: machtype=lemote-yeeloong-2f-7inch + make-printk-non-secret + Force pointers printed to the console to be unhashed. + By default, when a pointer is printed to the kernel + console (via %p format string), that pointer is + "hashed", i.e. obscured by hashing the pointer value. + This is a security feature that hides actual kernel + addresses from unprivileged users, but it also makes + debugging the kernel more difficult since unequal + pointers can no longer be compared. If this option is + specified, then all normal pointers will have their + true value printed. Pointers printed via %pK may + still be hashed. This option should only be specified + when debugging the kernel. Please do not use on + production kernels. + max_addr=nn[KMG] [KNL,BOOT,ia64] All physical memory greater than or equal to this physical address is ignored.lib/test_printf.c +8
-0@@ -35,6 +35,8 @@ KSTM_MODULE_GLOBALS();static char *test_buffer__initdata ;static char *alloced_buffer__initdata ; +extern bool debug_never_hash_pointers; + static int __printf(4, 0) __init do_test(int bufsize, const char *expect, int elen, const char *fmt, va_list ap)@@ -301,6 +303,12 @@ plain(void) {int err; +if (debug_never_hash_pointers) { +pr_warn ("skipping plain 'p' tests" ); + skipped_tests +=2 ; +return ; + } + err =plain_hash ();if (err) {pr_warn ("plain 'p' does not appear to be hashed\n" );lib/vsprintf.c +17
-0@@ -2090,6 +2090,34 @@ char *fwnode_string(char *buf, char *end, struct fwnode_handle *fwnode,return widen_string (buf, buf - buf_start, end, spec); } +/* Disable pointer hashing if requested */ +bool debug_never_hash_pointers__ro_after_init ; +EXPORT_SYMBOL_GPL (debug_never_hash_pointers); + +static int __init debug_never_hash_pointers_enable (char *str) +{ + debug_never_hash_pointers = true; + +pr_warn ("**********************************************************\n" ); +pr_warn ("** NOTICE NOTICE NOTICE NOTICE NOTICE NOTICE NOTICE **\n" ); +pr_warn ("** **\n" ); +pr_warn ("** All pointers that are printed to the console will **\n" ); +pr_warn ("** be printed as unhashed. **\n" ); +pr_warn ("** **\n" ); +pr_warn ("** Kernel memory addresses are exposed, which may **\n" ); +pr_warn ("** reduce the security of your system. **\n" ); +pr_warn ("** **\n" );
lib/vsprintf.c char *fwnode_string(char *buf, char *end, struct fwnode_handle *fwnode, · line 2119
Quoted 10 lines
lib/vsprintf.c +10
-0+pr_warn ("** If you see this message and you are not debugging **\n" ); +pr_warn ("** the kernel, report this immediately to your system **\n" ); +pr_warn ("** administrator! **\n" ); +pr_warn ("** **\n" ); +pr_warn ("** NOTICE NOTICE NOTICE NOTICE NOTICE NOTICE NOTICE **\n" ); +pr_warn ("**********************************************************\n" ); + +return 0 ; +} +early_param ("make-printk-non-secret" , debug_never_hash_pointers_enable);
The mail about EFENCE, and the different names of the parameter and the variable, made realized one important thing. The kernel parameter and the message does not describe the full effect.
The option causes that vsprintf() will not hash pointers. Yes, it is primary used by printk(). But it is used also in some other interfaces, especially trace_printk(), seq_buf() API. The naked pointers might appear more or less anywhere, including procfs, sysfs, debugfs.
IMHO, we should fix this. The long discussion was about how to make this option safe. Users should be aware that it is not only about the kernel log.
I suggest to rename the parameter "debug_never_hash_pointer" and use the same name for the parameter and the variable.
We also should make the warning more generic. I suggest to replace the first paragraph with something like:
pr_warn("** The hashing of printed pointers has been disabled **\n");
pr_warn("** for debugging purposes. **\n");Feel free to use a better wording. I am not a native speaker.
Of course, also kernel-parameters.txt has to be updated accordingly.
Best Regards,
Petr
Re: [PATCH 3/3] lib/vsprintf: make-printk-non-secret printks all addresses as unhashed
Quoted 6 lines · Petr Mladek
The option causes that
vsprintf()will not hash pointers. Yes, it is primary used byprintk(). But it is used also in some other interfaces, especiallytrace_printk(),seq_buf()API. The naked pointers might appear more or less anywhere, including procfs, sysfs, debugfs.
Fair point. Shouldn't calls to seq_buf_printf() (and any printk usage that always exists in the context of a user-space process) use %pK anyway?
Hmmm.... maybe vsprintf() should automatically replace %p with %pK if it detects a user-space context?
Quoted 3 lines
IMHO, we should fix this. The long discussion was about how to make this option safe. Users should be aware that it is not only about the kernel log.
Agreed.
Quoted 2 lines
I suggest to rename the parameter "
debug_never_hash_pointer" and use the same name for the parameter and the variable.
Will do.
Quoted 7 lines
We also should make the warning more generic. I suggest to replace the first paragraph with something like:
pr_warn("** The hashing of printed pointers has been disabled **\n"); pr_warn("** for debugging purposes. **\n");Feel free to use a better wording. I am not a native speaker.
You could have fooled me.
Quoted 1 line
Of course, also kernel-parameters.txt has to be updated accordingly.
Ok.
Petr Mladek
maintainerRe: [PATCH 3/3] lib/vsprintf: make-printk-non-secret printks all addresses as unhashed
Hi,
I have realized that I did not comment the two ideas.
Quoted 4 lines · Timur Tabi
Quoted 1 line · Petr Mladek
Quoted 8 lines
Quoted 5 lines
The option causes that
vsprintf()will not hash pointers. Yes, it is primary used byprintk(). But it is used also in some other interfaces, especiallytrace_printk(),seq_buf()API. The naked pointers might appear more or less anywhere, including procfs, sysfs, debugfs.Fair point. Shouldn't calls to
seq_buf_printf()(and any printk usage that always exists in the context of a user-space process) use %pK anyway?
seq_buf is a handy API that might be used for different purpose. For example, it seems to be used ftrace where people might want to see real pointers when debugging.
Quoted 2 lines
Hmmm.... maybe
vsprintf()should automatically replace %p with %pK if it detects a user-space context?
I am not sure if there is an easy and reliable way how to detect the user-space context. On some architectures, it might be possible to guess it by the address of the buffer. But it will not work when the message is temporary printed into a local buffer and copied later.
Let's keep it simple. Heuristics often become very complex over time.
Best Regards,
Petr
Quoted 29 lines · Timur Tabi
[accidentally sent from the wrong email address, so resending]
[The list of email addresses on CC: is getting quite lengthy, so I hope I've included everyone.]
Although hashing addresses printed via printk does make the kernel more secure, it interferes with debugging, especially with some functions like
print_hex_dump()which always uses hashed addresses.To avoid having to choose between %p and %px, it's easier to add a kernel command line that treats all %p as %px. This encourages developers to use %p more without making debugging more difficult.
Patches #1 and #2 upgrade the kselftest framework so that it can report on tests that were skipped outright. This is needed for the
test_printfmodule which will now skip %p hashing tests if hashing is disabled.Patch #2 upgrades the printf library to check the command line. It also updates
test_printf().Timur Tabi (3):
lib/test_printf: useKSTM_MODULE_GLOBALSmacro
kselftest: add support for skipped tests
[v2] lib/vsprintf: make-printk-non-secret printks all addresses as
unhashed
Quoted 5 lines
.../admin-guide/kernel-parameters.txt | 15 +++++++ lib/test_printf.c | 12 +++++- lib/vsprintf.c | 40 ++++++++++++++++++- tools/testing/selftests/kselftest_module.h | 18 ++++++--- 4 files changed, 75 insertions(+), 10 deletions(-)
I wanted to test this for deciding if we can show sensitive info in KFENCE reports, which works just fine now that debug_never_hash_pointers is non-static. FWIW,
But unfortunately this series broke some other test:
| In file included from lib/test_bitmap.c:17: | lib/test_bitmap.c: In function ‘test_bitmap_init’: | lib/../tools/testing/selftests/kselftest_module.h:45:48: error: ‘skipped_tests’ undeclared (first use in this function); did you mean ‘failed_tests’? | 45 | return kstm_report(total_tests, failed_tests, skipped_tests); \ | | ^~~~~~~~~~~~~ | lib/test_bitmap.c:637:1: note: in expansion of macro ‘KSTM_MODULE_LOADERS’ | 637 | KSTM_MODULE_LOADERS(test_bitmap); | | ^~~~~~~~~~~~~~~~~~~ | lib/../tools/testing/selftests/kselftest_module.h:45:48: note: each undeclared identifier is reported only once for each function it appears in | 45 | return kstm_report(total_tests, failed_tests, skipped_tests); \ | | ^~~~~~~~~~~~~ | lib/test_bitmap.c:637:1: note: in expansion of macro ‘KSTM_MODULE_LOADERS’ | 637 | KSTM_MODULE_LOADERS(test_bitmap); | | ^~~~~~~~~~~~~~~~~~~ | lib/../tools/testing/selftests/kselftest_module.h:46:1: error: control reaches end of non-void function [-Werror=return-type] | 46 | } \ | | ^ | lib/test_bitmap.c:637:1: note: in expansion of macro ‘KSTM_MODULE_LOADERS’ | 637 | KSTM_MODULE_LOADERS(test_bitmap); | | ^~~~~~~~~~~~~~~~~~~
My allyesconfig build suggests test_bitmap.c is the only one, so it should probably be fixed up in this series.
Thanks,
-- Marco
Quoted 5 lines · Marco Elver
I wanted to test this for deciding if we can show sensitive info in KFENCE reports, which works just fine now that
debug_never_hash_pointersis non-static. FWIW,Acked-by Marco Elver<[email protected]>
Thank you.
Quoted 5 lines
But unfortunately this series broke some other test:
| In file included from lib/test_bitmap.c:17: | lib/test_bitmap.c: In function ‘test_bitmap_init’: | lib/../tools/testing/selftests/kselftest_module.h:45:48: error: ‘skipped_tests’ undeclared (first use in this function); did you mean ‘failed_tests’?
This will be fixed in v3. Thanks.
Quoted 16 lines · Timur Tabi
[accidentally sent from the wrong email address, so resending]
[The list of email addresses on CC: is getting quite lengthy, so I hope I've included everyone.]
Although hashing addresses printed via printk does make the kernel more secure, it interferes with debugging, especially with some functions like
print_hex_dump()which always uses hashed addresses.To avoid having to choose between %p and %px, it's easier to add a kernel command line that treats all %p as %px. This encourages developers to use %p more without making debugging more difficult.
Quoted 7 lines
Patches #1 and #2 upgrade the kselftest framework so that it can report on tests that were skipped outright. This is needed for the
test_printfmodule which will now skip %p hashing tests if hashing is disabled.Patch #2 upgrades the printf library to check the command line. It also updates
test_printf().
It's a bit hard in some mailers (like Gmail) to see the different versions of your patches.
Can you use in the future
- either git format-patch -v<N> ..., where <N> is a version
- or git format-patch --subject-prefix="PATCH vX / RESEND / etc" ...
?
Signature · 2 lines
With Best Regards, Andy Shevchenko
Quoted 6 lines · Andy Shevchenko
It's a bit hard in some mailers (like Gmail) to see the different versions of your patches.
Can you use in the future
- eithergit format-patch -v<N> ..., where <N> is a version
- orgit format-patch --subject-prefix="PATCH vX / RESEND / etc" ...
?
Yes, of course. Sorry about that. Like I said earlier, I really shouldn't be posting patches late at night, when I will just forget important details.
Quoted 9 lines · Timur Tabi
[accidentally sent from the wrong email address, so resending]
[The list of email addresses on CC: is getting quite lengthy, so I hope I've included everyone.]
Although hashing addresses printed via printk does make the kernel more secure, it interferes with debugging, especially with some functions like
print_hex_dump()which always uses hashed addresses.
Oh, I was wishing
lib/vsprintf.c +1 -1
@@ -802,7 +802,7 @@ static char *ptr_to_id(char *buf, char *end, const void *ptr,* Print the real pointer value for NULL and error pointers, * as they are not actual addresses. */ -if (IS_ERR_OR_NULL (ptr)) +if (IS_ERR_OR_NULL (ptr) ||IS_ENABLED (CONFIG_DEBUG_DONT_HASH_POINTERS ))return pointer_string (buf, end, ptr, spec);/* When debugging early boot use non-cryptographically secure hash. */
change as a kernel config option, for more we try to switch using kernel command line options, more we likely make errors with sharing appropriate kernel command line options (e.g. https://github.com/google/syzkaller/commit/99c64d5c672700d6c0de63d11db25a0678e47a75 ).
Steven Rostedt
maintainerQuoted 16 lines · On Thu, 11 Feb 2021 00:46:15 +0900 Tetsuo Handa…
Oh, I was wishing
lib/vsprintf.c +1
-1@@ -802,7 +802,7 @@ static char *ptr_to_id(char *buf, char *end, const void *ptr,* Print the real pointer value for NULL and error pointers, * as they are not actual addresses. */ -if (IS_ERR_OR_NULL (ptr)) +if (IS_ERR_OR_NULL (ptr) ||IS_ENABLED (CONFIG_DEBUG_DONT_HASH_POINTERS ))return pointer_string (buf, end, ptr, spec);/* When debugging early boot use non-cryptographically secure hash. */
Quoted 3 lines
change as a kernel config option, for more we try to switch using kernel command line options, more we likely make errors with sharing appropriate kernel command line options (e.g. https://github.com/google/syzkaller/commit/99c64d5c672700d6c0de63d11db25a0678e47a75 ).
The entire point of this exercise is not to make it easy to add this feature. Linus was absolutely against a config option, and I am too.
The point of this exercise is to be able to debug the *same* kernel that someone is having issues with. And this is to facilitate that debugging. Whereas the example you show, the command line modifies how the kernel works. This command line only modifies what the kernel displays. Big difference.
-- Steve
Quoted 2 lines · Steven Rostedt
The point of this exercise is to be able to debug the *same* kernel that someone is having issues with. And this is to facilitate that debugging.
That's too difficult to use. If a problem is not reproducible, we will have no choice but always specify "never hash pointers" command line option. If a problem is reproducible, we can rebuild that kernel with "never hash pointers" config option turned on.
Steven Rostedt
maintainerQuoted 8 lines · On Thu, 11 Feb 2021 01:39:41 +0900 Tetsuo Handa…
Quoted 2 lines · Steven Rostedt
The point of this exercise is to be able to debug the *same* kernel that someone is having issues with. And this is to facilitate that debugging.
That's too difficult to use. If a problem is not reproducible, we will have no choice but always specify "never hash pointers" command line option. If a problem is reproducible, we can rebuild that kernel with "never hash pointers" config option turned on.
Now the question is, why do you need the unhashed pointer?
Currently, the instruction pointer is what is fine right? You get the a function and its offset. If there's something that is needed, perhaps we should look at how to fix that, instead of just unhashing all pointers by default.
-- Steve
Quoted 7 lines · Steven Rostedt
Quoted 4 lines · On Thu, 11 Feb 2021 01:39:41 +0900 Tetsuo Handa…
Quoted 2 lines · Steven Rostedt
The point of this exercise is to be able to debug the *same* kernel that someone is having issues with. And this is to facilitate that debugging.
Quoted 6 lines
Quoted 4 lines
That's too difficult to use. If a problem is not reproducible, we will have no choice but always specify "never hash pointers" command line option. If a problem is reproducible, we can rebuild that kernel with "never hash pointers" config option turned on.
Now the question is, why do you need the unhashed pointer?
Because unhashed pointers might give some clue. We can rebuild the same kernel using the same kernel config / compiler etc. and compare unhashed pointers with addresses in System.map / kallsyms files without reproducing the problem.
Quoted 5 lines
Currently, the instruction pointer is what is fine right? You get the a function and its offset. If there's something that is needed, perhaps we should look at how to fix that, instead of just unhashing all pointers by default.
I'm not refusing to use kernel command line options. I'm expecting that we can also hardcode using kernel config options. Since boot-time switching via kernel command line options makes the kernel behave differently, less boot-time switching is better for avoiding unexpected problems (e.g. unintended LSM was enabled).
Steven Rostedt
maintainerQuoted 5 lines · On Thu, 11 Feb 2021 02:07:21 +0900 Tetsuo Handa…
I'm not refusing to use kernel command line options. I'm expecting that we can also hardcode using kernel config options. Since boot-time switching via kernel command line options makes the kernel behave differently, less boot-time switching is better for avoiding unexpected problems (e.g. unintended LSM was enabled).
You could just add:
CONFIG_CMDLINE_BOOL=y
CONFIG_CMDLINE="make-printk-non-secret"
If you want it part of your config.
-- Steve
Quoted 6 lines · Steven Rostedt
Now the question is, why do you need the unhashed pointer?
Currently, the instruction pointer is what is fine right? You get the a function and its offset. If there's something that is needed, perhaps we should look at how to fix that, instead of just unhashing all pointers by default.
The original version of this patch only fixed print_hex_dump(), because hashed addresses didn't make any sense for that. Each address is incremented by 16 or 32, but since they were all hashed, they may as well have been random numbers.
Andy Shevchenko
reviewerQuoted 8 lines · Tetsuo Handa
Quoted 2 lines · Steven Rostedt
The point of this exercise is to be able to debug the *same* kernel that someone is having issues with. And this is to facilitate that debugging.
That's too difficult to use. If a problem is not reproducible, we will have no choice but always specify "never hash pointers" command line option. If a problem is reproducible, we can rebuild that kernel with "never hash pointers" config option turned on.
I think what you are targeting is something like dynamic debug approach where you can choose which prints to enable/disable and what enable/disable in them.
In that case you specifically apply a command line option and enable only files / lines in the files.
Signature · 2 lines
With Best Regards, Andy Shevchenko
Quoted 5 lines · Andy Shevchenko
Quoted 4 lines · Tetsuo Handa
Quoted 2 lines · Steven Rostedt
The point of this exercise is to be able to debug the *same* kernel that someone is having issues with. And this is to facilitate that debugging.
Quoted 7 lines
Quoted 4 lines
That's too difficult to use. If a problem is not reproducible, we will have no choice but always specify "never hash pointers" command line option. If a problem is reproducible, we can rebuild that kernel with "never hash pointers" config option turned on.
I think what you are targeting is something like dynamic debug approach where you can choose which prints to enable/disable and what enable/disable in them.
What I'm targeting is "zero interaction from kernel command line options" like
https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/commit/drivers/usb/usbip?id=f1bdf414e7dd0cbc26460425719fc3ea479947a2 .
Quoted 3 lines
In that case you specifically apply a command line option and enable only files / lines in the files.
While there is boot-config feature for specifying very long kernel command line options, I can't enforce syzkaller users (including syzbot) to switch what to enable/disable via kernel command line options. Let alone defining a kernel command line option for single-purpose debug printk() changes like shown above.