2015-12-22 22:52:39

by Dave Hansen

[permalink] [raw]
Subject: [PATCH 0/5] x86: early command-line parsing fixes / tests (v2)

The code here is unchanged except for fixing a compile error that
I managed to introduce just before I sent it last time.

The tests for this code are in the last two patches. If folks
think it is overkill, those can be left out. But, the first
three are still needed.

Cc: Borislav Petkov <[email protected]>
Cc: H. Peter Anvin <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]


2015-12-22 22:52:41

by Dave Hansen

[permalink] [raw]
Subject: [PATCH 1/5] x86: fix early command-line parsing when matching at end


From: Dave Hansen <[email protected]>

The x86 early command line parsing in cmdline_find_option_bool()
is buggy. If it matches a specified 'option' all the way to
the end of the command-line, it will consider it a match.

For instance,

cmdline = "foo";
cmdline_find_option_bool(cmdline, "fool");

will return 1. This is particularly annoying since we have
actual FPU options like "noxsave" and "noxsaves" So,
command-line "foo bar noxsave" will match *BOTH* a "noxsave" and
"noxsaves". (This turns out not to be an actual problem because
"noxsave" implies "noxsaves", but it's still confusing.)

To fix this, we simplify the code and stop tracking 'len'. 'len'
was trying to indicate either the NULL terminator *OR* the end of
a non-NULL-terminated commandline at 'COMMAND_LINE_SIZE'. But,
each of the three states is *already* checking 'cmdline' for a
NULL terminator.

We _only_ need to check if we have overrun 'COMMAND_LINE_SIZE',
and that we can do without keeping 'len' around.

Also add some commends to clarify what is going on.

Signed-off-by: Dave Hansen <[email protected]>
Cc: Borislav Petkov <[email protected]>
Cc: H. Peter Anvin <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
---

b/arch/x86/lib/cmdline.c | 34 ++++++++++++++++++++++++----------
1 file changed, 24 insertions(+), 10 deletions(-)

diff -puN arch/x86/lib/cmdline.c~x86-broken-end-of-string-command-line-parsing arch/x86/lib/cmdline.c
--- a/arch/x86/lib/cmdline.c~x86-broken-end-of-string-command-line-parsing 2015-12-22 11:56:58.639149442 -0800
+++ b/arch/x86/lib/cmdline.c 2015-12-22 11:56:58.642149577 -0800
@@ -21,12 +21,14 @@ static inline int myisspace(u8 c)
* @option: option string to look for
*
* Returns the position of that @option (starts counting with 1)
- * or 0 on not found.
+ * or 0 on not found. @option will only be found if it is found
+ * as an entire word in @cmdline. For instance, if @option="car"
+ * then a cmdline which contains "cart" will not match.
*/
int cmdline_find_option_bool(const char *cmdline, const char *option)
{
char c;
- int len, pos = 0, wstart = 0;
+ int pos = 0, wstart = 0;
const char *opptr = NULL;
enum {
st_wordstart = 0, /* Start of word/after whitespace */
@@ -37,11 +39,14 @@ int cmdline_find_option_bool(const char
if (!cmdline)
return -1; /* No command line */

- len = min_t(int, strlen(cmdline), COMMAND_LINE_SIZE);
- if (!len)
+ if (!strlen(cmdline))
return 0;

- while (len--) {
+ /*
+ * This 'pos' check ensures we do not overrun
+ * a non-NULL-terminated 'cmdline'
+ */
+ while (pos < COMMAND_LINE_SIZE) {
c = *(char *)cmdline++;
pos++;

@@ -58,17 +63,26 @@ int cmdline_find_option_bool(const char
/* fall through */

case st_wordcmp:
- if (!*opptr)
+ if (!*opptr) {
+ /*
+ * We matched all the way to the end of the
+ * option we were looking for. If the
+ * command-line has a space _or_ ends, then
+ * we matched!
+ */
if (!c || myisspace(c))
return wstart;
else
state = st_wordskip;
- else if (!c)
+ } else if (!c) {
+ /*
+ * Hit the NULL terminator on the end of
+ * cmdline.
+ */
return 0;
- else if (c != *opptr++)
+ } else if (c != *opptr++) {
state = st_wordskip;
- else if (!len) /* last word and is matching */
- return wstart;
+ }
break;

case st_wordskip:
_

2015-12-22 22:53:36

by Dave Hansen

[permalink] [raw]
Subject: [PATCH 2/5] x86: fix early command-line parsing, when partial word match


From: Dave Hansen <[email protected]>

cmdline_find_option_bool() keeps track of position in two strings:
1. the command-line
2. the option we are searchign for in the command-line

We plow through each character in the command-line one at a time,
always moving forward. We move forward in the option ('opptr')
when we match characters in 'cmdline'. We reset the 'opptr' only
when we go in to the 'st_wordstart' state.

But, if we fail to match an option because we see a space (
state=st_wordcmp, *opptr='\0',c=' '), we set state='st_wordskip'
and 'break', moving to the next character. But, that move to
the next character is the one *after* the ' '. This means that
we will miss a 'st_wordstart' state.

For instance, if we have

cmdline = "foo fool";

and are searching for "fool", we have:

"fool"
opptr = ----^

"foo fool"
c = --------^

We see that 'l' != ' ', set state=st_wordskip, break, and then
move 'c', so:

"foo fool"
c = ---------^

and are still in state=st_wordskip. We will stay in wordskip
until we have skipped "fool", thus missing the option we were
looking for. This *only* happens when you have a partially-
matching word followed by a matching one.

To fix this, we always fall *into* the 'st_wordskip' state when
we set it.

Signed-off-by: Dave Hansen <[email protected]>
Cc: Borislav Petkov <[email protected]>
Cc: H. Peter Anvin <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
---

b/arch/x86/lib/cmdline.c | 18 +++++++++++++-----
1 file changed, 13 insertions(+), 5 deletions(-)

diff -puN arch/x86/lib/cmdline.c~x86-mid-option-match-command-line-parsing arch/x86/lib/cmdline.c
--- a/arch/x86/lib/cmdline.c~x86-mid-option-match-command-line-parsing 2015-12-22 11:56:59.047167827 -0800
+++ b/arch/x86/lib/cmdline.c 2015-12-22 11:56:59.050167962 -0800
@@ -72,18 +72,26 @@ int cmdline_find_option_bool(const char
*/
if (!c || myisspace(c))
return wstart;
- else
- state = st_wordskip;
+ /*
+ * We hit the end of the option, but _not_
+ * the end of a word on the cmdline. Not
+ * a match.
+ */
} else if (!c) {
/*
* Hit the NULL terminator on the end of
* cmdline.
*/
return 0;
- } else if (c != *opptr++) {
- state = st_wordskip;
+ } else if (c == *opptr++) {
+ /*
+ * We are currently matching, so continue
+ * to the next character on the cmdline.
+ */
+ break;
}
- break;
+ state = st_wordskip;
+ /* fall through */

case st_wordskip:
if (!c)
_

2015-12-22 22:53:16

by Dave Hansen

[permalink] [raw]
Subject: [PATCH 3/5] x86: simplify early command line parsing


From: Dave Hansen <[email protected]>

__cmdline_find_option_bool() tries to account for both
NULL-terminated and non-NULL-terminated strings. It keeps 'pos'
to look for the end of the buffer and also looks for '!c' in a
bunch of places to look for NULL termination.

But, it also calls strlen(). You can't call strlen on a
non-NULL-terminated string.

If !strlen(cmdline), then cmdline[0]=='\0'. In that case, we
will go in to the while() loop, set c='\0', hit st_wordstart,
notice !c, and will immediately return 0.

So, remove the strlen(). It is unnecessary and unsafe.

Signed-off-by: Dave Hansen <[email protected]>
Cc: Borislav Petkov <[email protected]>
Cc: H. Peter Anvin <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
---

b/arch/x86/lib/cmdline.c | 3 ---
1 file changed, 3 deletions(-)

diff -puN arch/x86/lib/cmdline.c~x86-early-command-line-non-term arch/x86/lib/cmdline.c
--- a/arch/x86/lib/cmdline.c~x86-early-command-line-non-term 2015-12-22 11:56:59.454186167 -0800
+++ b/arch/x86/lib/cmdline.c 2015-12-22 11:56:59.457186302 -0800
@@ -39,9 +39,6 @@ int cmdline_find_option_bool(const char
if (!cmdline)
return -1; /* No command line */

- if (!strlen(cmdline))
- return 0;
-
/*
* This 'pos' check ensures we do not overrun
* a non-NULL-terminated 'cmdline'
_

2015-12-22 22:52:46

by Dave Hansen

[permalink] [raw]
Subject: [PATCH 4/5] x86: pass in size to early cmdline parsing


From: Dave Hansen <[email protected]>

We will use this in a few patches to implement tests for early
parsing.

Signed-off-by: Dave Hansen <[email protected]>
Cc: Borislav Petkov <[email protected]>
Cc: H. Peter Anvin <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
---

b/arch/x86/lib/cmdline.c | 11 +++++++++--
1 file changed, 9 insertions(+), 2 deletions(-)

diff -puN arch/x86/lib/cmdline.c~pass-in-cmdline-size arch/x86/lib/cmdline.c
--- a/arch/x86/lib/cmdline.c~pass-in-cmdline-size 2015-12-22 11:56:59.859204417 -0800
+++ b/arch/x86/lib/cmdline.c 2015-12-22 11:56:59.862204552 -0800
@@ -25,7 +25,8 @@ static inline int myisspace(u8 c)
* as an entire word in @cmdline. For instance, if @option="car"
* then a cmdline which contains "cart" will not match.
*/
-int cmdline_find_option_bool(const char *cmdline, const char *option)
+static int __cmdline_find_option_bool(const char *cmdline,
+ int max_cmdline_size, const char *option)
{
char c;
int pos = 0, wstart = 0;
@@ -43,7 +44,7 @@ int cmdline_find_option_bool(const char
* This 'pos' check ensures we do not overrun
* a non-NULL-terminated 'cmdline'
*/
- while (pos < COMMAND_LINE_SIZE) {
+ while (pos < max_cmdline_size) {
c = *(char *)cmdline++;
pos++;

@@ -101,3 +102,9 @@ int cmdline_find_option_bool(const char

return 0; /* Buffer overrun */
}
+
+int cmdline_find_option_bool(const char *cmdline, const char *option)
+{
+ return __cmdline_find_option_bool(cmdline, COMMAND_LINE_SIZE,
+ option);
+}
_

2015-12-22 22:52:56

by Dave Hansen

[permalink] [raw]
Subject: [PATCH 5/5] x86: test early command-line code


From: Dave Hansen <[email protected]>

Here are some simple tests for the early command-line code. It
had way more bugs than it should have, so let's make sure they
never pop up again.

Note, there are a few failures in here now. We will fix those
up in the next few patches.

This is complete overkill for this code, but I had to do it to
convince myself that I wasn't making it any worse.

Signed-off-by: Dave Hansen <[email protected]>
Cc: Borislav Petkov <[email protected]>
Cc: H. Peter Anvin <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
---

b/arch/x86/Kconfig.debug | 7 ++
b/arch/x86/kernel/check.c | 1
b/arch/x86/kernel/setup.c | 9 +++
b/arch/x86/lib/cmdline.c | 122 ++++++++++++++++++++++++++++++++++++++++++++++
4 files changed, 138 insertions(+), 1 deletion(-)

diff -puN arch/x86/Kconfig.debug~x86-early-command-line-test arch/x86/Kconfig.debug
--- a/arch/x86/Kconfig.debug~x86-early-command-line-test 2015-12-22 11:57:01.229266152 -0800
+++ b/arch/x86/Kconfig.debug 2015-12-22 11:57:01.238266557 -0800
@@ -400,4 +400,11 @@ config PUNIT_ATOM_DEBUG
The current power state can be read from
/sys/kernel/debug/punit_atom/dev_power_state

+config X86_TEST_EARLY_CMDLINE
+ bool "Early command line runtime tests"
+ ---help---
+ This creates some test command-lines and tries to parse
+ a bunch of options from them. The overhead is small both
+ at boot and increased text/data sizes.
+
endmenu
diff -puN arch/x86/kernel/check.c~x86-early-command-line-test arch/x86/kernel/check.c
--- a/arch/x86/kernel/check.c~x86-early-command-line-test 2015-12-22 11:57:01.230266197 -0800
+++ b/arch/x86/kernel/check.c 2015-12-22 11:57:01.239266602 -0800
@@ -164,4 +164,3 @@ static int start_periodic_check_for_corr
return 0;
}
device_initcall(start_periodic_check_for_corruption);
-
diff -puN arch/x86/lib/cmdline.c~x86-early-command-line-test arch/x86/lib/cmdline.c
--- a/arch/x86/lib/cmdline.c~x86-early-command-line-test 2015-12-22 11:57:01.232266287 -0800
+++ b/arch/x86/lib/cmdline.c 2015-12-22 11:57:01.239266602 -0800
@@ -4,6 +4,7 @@
*
* Misc librarized functions for cmdline poking.
*/
+#include <linux/bug.h>
#include <linux/kernel.h>
#include <linux/string.h>
#include <linux/ctype.h>
@@ -108,3 +109,124 @@ int cmdline_find_option_bool(const char
return __cmdline_find_option_bool(cmdline, COMMAND_LINE_SIZE,
option);
}
+
+#ifdef CONFIG_X86_TEST_EARLY_CMDLINE
+
+static int __cmdtest(char *cmdline, int str_size, char *option,
+ int expected_result, int do_shrink)
+{
+ int ret;
+ int null_terminate;
+ /* Results are 1-based, so bias back down by 1 */
+ int option_end = expected_result + strlen(option) - 1;
+ int shrink_max = 0;
+
+ if (cmdline && do_shrink)
+ shrink_max = strlen(cmdline);
+ /*
+ * The option was not found. If it was not found in the
+ * *full* command-line, it should never be found in any
+ * *part* of the command-line.
+ */
+ for (null_terminate = 0; null_terminate <= 1; null_terminate++) {
+ int shrink_by;
+ for (shrink_by = 0; shrink_by < shrink_max; shrink_by++) {
+ int str_size_tst = str_size - shrink_by;
+ char tmp = cmdline[str_size_tst];
+
+ /*
+ * Do not run tests that would truncate
+ * over the expected option
+ */
+ if (str_size_tst <= option_end)
+ continue;
+
+ if (null_terminate)
+ cmdline[str_size_tst] = '\0';
+ ret = __cmdline_find_option_bool(cmdline, str_size_tst,
+ option);
+ if (null_terminate)
+ cmdline[str_size_tst] = tmp;
+
+ if (ret == expected_result)
+ continue;
+ pr_err("failed cmdline test ('%s', %d, '%s') == %d "
+ "nulld: %d got: %d\n",
+ cmdline, str_size_tst, option,
+ expected_result, null_terminate,
+ ret);
+ return 1;
+ }
+ }
+ return 0;
+}
+
+#define cmdtest(cmdline, option, result) \
+ WARN_ON(__cmdtest(cmdline, sizeof(cmdline), option, result, 1))
+
+#define cmdtest_noshrink(cmdline, option, result) \
+ WARN_ON(__cmdtest(cmdline, sizeof(cmdline), option, result, 0))
+
+char cmdline1[] = "CALL me Ishmael ";
+char cmdline2[] = "Whenever I find myself growing grim about the mouth ";
+char cmdline3[] = "grow growing ";
+int test_early_cmdline(void)
+{
+ /* NULL command-line: */
+ WARN_ON(__cmdline_find_option_bool(NULL, 22, "Ishmael") != -1);
+ /* zero-length command-line: */
+ cmdtest("", "Ishmael", 0);
+
+ /* Find words at each of 3 positions: start, middle, end */
+ cmdtest(cmdline1, "CALL", 1);
+ cmdtest(cmdline1, "me", 6);
+ cmdtest(cmdline1, "Ishmael", 9);
+
+ /*
+ * Fail to find strings that all occur in the cmdline,
+ * but not as full words
+ */
+ /*
+ * If "option" is _present_ in "cmdline" as the start of a
+ * word, like cmdline="foo bar" and we pass in option="b",
+ * when we shrink cmdline to "foo b", it will match. So,
+ * skip shrink tests for those.
+ */
+ cmdtest_noshrink(cmdline1, "m", 0);
+ cmdtest(cmdline1, "e", 0);
+ cmdtest(cmdline1, "C", 0);
+ cmdtest(cmdline1, "l", 0);
+ cmdtest_noshrink(cmdline1, "Ishmae", 0);
+ cmdtest(cmdline1, "mael", 0);
+ /*
+ * Look for strings that do not occur, but match until
+ * close to the end of cmdline
+ */
+ cmdtest_noshrink(cmdline1, "Ishmae", 0);
+ cmdtest(cmdline1, "Ishmaels", 0);
+ cmdtest(cmdline1, "maels", 0);
+
+ /*
+ * Look for full words that do not occur in a different
+ * cmdline
+ */
+ cmdtest(cmdline2, "CALL", 0);
+ cmdtest(cmdline2, "me", 0);
+ cmdtest(cmdline2, "Ishmael", 0);
+ /*
+ * Look for full words which do occur in cmdline2
+ */
+ cmdtest(cmdline2, "Whenever", 1);
+ cmdtest(cmdline2, "growing", 24);
+ cmdtest(cmdline2, "grim", 32);
+ cmdtest(cmdline2, "mouth", 47);
+
+ /*
+ * Catch the bug where if we match a partial word and
+ * then have a space, we do not match the _next_ word.
+ */
+ cmdtest(cmdline3, "grow", 1);
+ cmdtest(cmdline3, "growing", 6);
+ return 0;
+}
+#endif /* CONFIG_X86_TEST_EARLY_CMDLINE */
diff -puN arch/x86/kernel/setup.c~x86-early-command-line-test arch/x86/kernel/setup.c
--- a/arch/x86/kernel/setup.c~x86-early-command-line-test 2015-12-22 11:57:01.233266332 -0800
+++ b/arch/x86/kernel/setup.c 2015-12-22 11:57:01.239266602 -0800
@@ -1282,3 +1282,12 @@ static int __init register_kernel_offset
return 0;
}
__initcall(register_kernel_offset_dumper);
+
+/*
+ * This code is in lib/ and we do not link initcalls from there.
+ * Stash it here instead.
+ */
+#ifdef CONFIG_X86_TEST_EARLY_CMDLINE
+int test_early_cmdline(void);
+late_initcall(test_early_cmdline);
+#endif /* CONFIG_X86_TEST_EARLY_CMDLINE */
_