Skip to contentSkip to conversation

EDAC/mc: Prefer strscpy or scnprintf over strcpy

PATCH In review Len Baker 4 participants 8 messages Sep 3, 2021 – Sep 18, 2021 EDAC-core

lore
Share & tools
lore

Len Baker

author
#0 lore raw

PATCH v6 1/1 EDAC/mc: Prefer strscpy or scnprintf over strcpy

strcpy() performs no bounds checking on the destination buffer. This could result in linear overflows beyond the end of the buffer, leading to all kinds of misbehaviors. The safe replacement is strscpy() [1].

However, to simplify and clarify the code, to concatenate labels use the scnprintf() function. This way it is not necessary to check the return value of strscpy (-E2BIG if the parameter count is 0 or the src was truncated) since the scnprintf returns always the number of chars written into the buffer. This function returns always a nul-terminated string even if it needs to be truncated.

The main reason behind this patch is to remove all the strcpy() uses from the kernel with the purpose to clean up the proliferation of str*cpy() functions. Later on, the next step will be remove all the strcpy implementations [2].

[1] https://www.kernel.org/doc/html/latest/process/deprecated.html#strcpy
[2] https://github.com/KSPP/linux/issues/88

---
Hi Joe,

I have added the "Co-developed-by: Joe Perches" tag to give you credit, since all the code used in this patch relies heavily on your review code snippets.

I hope there are no objections.

Thanks,
Len

Changelog v1 -> v2
- Use the strscpy() instead of scnprintf() to add labels and follow a code pattern more similar to the current one (advance "p" and decrement "left") (Robert Richter).

Changelog v2 -> v3
- Rename the "left" variable to "len" (Robert Richter).
- Use strlen(p) instead of strlen(OTHER_LABEL) to decrement "len" and increment "p" as otherwise "left" could underflow and p overflow (Robert Richter).

Changelog v3 -> v4
- Change the commit subject (Joe Perches).
- Fix broken logic (Robert Richter).
- Rebase against v5.14-rc5.

Changelog v4 -> v5
- Change the commit subject.
- Clarify why the change is being made by adding more info to the commit message (Borislav Petkov).
- Use scnprintf instead of strscpy (Joe Perches).
- Rebase against v5.14-rc7.

Changelog v5 -> v6
- Rebase against v5.14.
- Refactor the code to use a more common scnprintf mechanism (Joe Perches).

Previous versions:

v1
https://lore.kernel.org/linux-hardening/[email protected]/

v2
https://lore.kernel.org/linux-hardening/[email protected]/

v3
https://lore.kernel.org/linux-hardening/[email protected]/

v4
https://lore.kernel.org/linux-hardening/[email protected]/

v5
https://lore.kernel.org/linux-hardening/[email protected]/

 drivers/edac/edac_mc.c | 15 +++++++--------
 1 file changed, 7 insertions(+), 8 deletions(-)
drivers/edac/edac_mc.c +7 -8
index f6d462d0be2d..97dff62970a5 100644
@@ -1032,6 +1032,8 @@ void edac_mc_handle_error(const enum hw_event_mc_err_type type,
 	int i, n_labels = 0;
 	struct edac_raw_error_desc *e = &mci->error_desc;
 	bool any_memory = true;
+	const char *prefix = "";
+	int n = 0;
 
 	edac_dbg(3, "MC%d\n", mci->mc_idx);
 
@@ -1113,12 +1115,9 @@ void edac_mc_handle_error(const enum hw_event_mc_err_type type,
 			p = e->label;
 			*p = '\0';
 		} else {
-			if (p != e->label) {
-				strcpy(p, OTHER_LABEL);
-				p += strlen(OTHER_LABEL);
-			}
-			strcpy(p, dimm->label);
-			p += strlen(p);
+			n += scnprintf(e->label + n, sizeof(e->label) - n,
+				       "%s%s", prefix, dimm->label);
+			prefix = OTHER_LABEL;
 		}
 
 		/*
@@ -1140,9 +1139,9 @@ void edac_mc_handle_error(const enum hw_event_mc_err_type type,
 	}
 
 	if (any_memory)
-		strcpy(e->label, "any memory");
+		strscpy(e->label, "any memory", sizeof(e->label));
 	else if (!*e->label)
-		strcpy(e->label, "unknown memory");
+		strscpy(e->label, "unknown memory", sizeof(e->label));
 
 	edac_inc_csrow(e, row, chan);
 

--
2.25.1

#1 lore raw
Quoted 1 line · Len Baker

strcpy() performs no bounds checking on the destination buffer. [email protected]/

[]

Quoted 4 lines
drivers/edac/edac_mc.c +0 -0
@@ -1113,12 +1115,9 @@ void edac_mc_handle_error(const enum
 hw_event_mc_err_type type,
 			p = e->label;
 			*p = '\0';

drivers/edac/edac_mc.c void edac_mc_handle_error(const enum · line 1121

Quoted 10 lines
drivers/edac/edac_mc.c +3 -6
 		} else {
-			if (p != e->label) {
-				strcpy(p, OTHER_LABEL);
-				p += strlen(OTHER_LABEL);
-			}
-			strcpy(p, dimm->label);
-			p += strlen(p);
+			n += scnprintf(e->label + n, sizeof(e->label) - n,
+				       "%s%s", prefix, dimm->label);
+			prefix = OTHER_LABEL;

OTHER_LABEL is a define specific to this module

IMO: Used once text macros are just obfuscating and should be removed.

author
#2 lore raw

Hi,

Quoted 23 lines · Joe Perches
Quoted 2 lines · Len Baker

strcpy() performs no bounds checking on the destination buffer. [email protected]/

[]

Quoted 4 lines
drivers/edac/edac_mc.c +0 -0
@@ -1113,12 +1115,9 @@ void edac_mc_handle_error(const enum
 hw_event_mc_err_type type,
 			p = e->label;
 			*p = '\0';

drivers/edac/edac_mc.c void edac_mc_handle_error(const enum · line 1121

Quoted 10 lines
drivers/edac/edac_mc.c +3 -6
 		} else {
-			if (p != e->label) {
-				strcpy(p, OTHER_LABEL);
-				p += strlen(OTHER_LABEL);
-			}
-			strcpy(p, dimm->label);
-			p += strlen(p);
+			n += scnprintf(e->label + n, sizeof(e->label) - n,
+				       "%s%s", prefix, dimm->label);
+			prefix = OTHER_LABEL;

OTHER_LABEL is a define specific to this module

Quoted 1 line

IMO: Used once text macros are just obfuscating and should be removed.

This macro is used in "/include/linux/edac.h" as follows:

struct edac_raw_error_desc {
	[...]
	char label[(EDAC_MC_LABEL_LEN + 1 + sizeof(OTHER_LABEL)) * EDAC_MAX_LABELS];
	[...]
};

If we remove this define the size of label would be:

	char label[(EDAC_MC_LABEL_LEN + 6) * EDAC_MAX_LABELS];

So, I think now is more complicated to understand because the size is what it is. If you prefer this option, I can remove the macro and add a comment with some explanation.

Regards,
Len

maintainer
#3 lore raw
Quoted 1 line · Len Baker

I can remove the macro and add a comment with some explanation.

No, please leave the macro.

Signature · 4 lines
Regards/Gruss,
    Boris.

https://people.kernel.org/tglx/notes-about-netiquette
author
#4 lore raw

Hi,

Quoted 4 lines · Borislav Petkov
Quoted 1 line · Len Baker

I can remove the macro and add a comment with some explanation.

No, please leave the macro.

Ok, so I leave the patch as is.

Thanks,
Len

#5 lore raw

PATCH 1/1 EDAC/mc: Prefer strscpy or scnprintf over strcpy, sprintf and snprintf

In reply to Len Baker

Len,

Quoted 16 lines · Len Baker

strcpy() performs no bounds checking on the destination buffer. This could result in linear overflows beyond the end of the buffer, leading to all kinds of misbehaviors. The safe replacement is strscpy() [1].

However, to simplify and clarify the code, to concatenate labels use the scnprintf() function. This way it is not necessary to check the return value of strscpy (-E2BIG if the parameter count is 0 or the src was truncated) since the scnprintf returns always the number of chars written into the buffer. This function returns always a nul-terminated string even if it needs to be truncated.

The main reason behind this patch is to remove all the strcpy() uses from the kernel with the purpose to clean up the proliferation of str*cpy() functions. Later on, the next step will be remove all the strcpy implementations [2].

Quoted 6 lines

[1] https://www.kernel.org/doc/html/latest/process/deprecated.html#strcpy
[2] https://github.com/KSPP/linux/issues/88

this patch looks good to me. I made some changes on top of it to further ease pointer arithmetic and also fix remaining sprintf/snprintf() users as it makes sense to have them all in a single change. See below. Boris, please apply.

Thanks,

-Robert

From 01a3c62a533e71984dfff7189e247b3e848f1449 Mon Sep 17 00:00:00 2001
From: Len Baker <[email protected]>
Date: Fri, 3 Sep 2021 17:05:39 +0200
Subject: [PATCH] EDAC/mc: Prefer strscpy or scnprintf over strcpy, sprintf
and snprintf

strcpy() performs no bounds checking on the destination buffer. This could result in linear overflows beyond the end of the buffer, leading to all kinds of misbehaviors. The safe replacement is strscpy(). [1][2]

However, to simplify and clarify the code, to concatenate labels use the scnprintf() function. This way it is not necessary to check the return value of strscpy (-E2BIG if the parameter count is 0 or the src was truncated) since the scnprintf returns always the number of chars written into the buffer. This function returns always a nul-terminated string even if it needs to be truncated.

While at it, fix all other broken string generation code that wrongly interprets snprintf()'s return code or just uses sprintf(), implement that using scnprintf() here too. Drop breaks in loops around scnprintf() as it is safe now to loop. Moreover, the check is
needless: For the case when the buffer is exhausted, len never gets zero because scnprintf() takes the full buffer length as input parameter, but excludes the trailing '\0' in its return code and thus, 1 is the minimum len.

[1] https://www.kernel.org/doc/html/latest/process/deprecated.html#strcpy
[2] https://github.com/KSPP/linux/issues/88

[ rric: Replace snprintf() with scnprintf(), rework sprintf() user,
drop breaks in loops around scnprintf(), introduce 'end' pointer to reduce pointer arithmetic, use prefix pattern for e->location, adjust subject and description ]

---
 drivers/edac/edac_mc.c | 42 ++++++++++++++++++------------------------
 1 file changed, 18 insertions(+), 24 deletions(-)
drivers/edac/edac_mc.c +18 -24
index 2c5975674723..9f82ca295353 100644
@@ -66,14 +66,12 @@ unsigned int edac_dimm_info_location(struct dimm_info *dimm, char *buf,
 	char *p = buf;
 
 	for (i = 0; i < mci->n_layers; i++) {
-		n = snprintf(p, len, "%s %d ",
+		n = scnprintf(p, len, "%s %d ",
 			      edac_layer_name[mci->layers[i].type],
 			      dimm->location[i]);
 		p += n;
 		len -= n;
 		count += n;
-		if (!len)
-			break;
 	}
 
 	return count;
@@ -341,19 +339,16 @@ static int edac_mc_alloc_dimms(struct mem_ctl_info *mci)
 		 */
 		len = sizeof(dimm->label);
 		p = dimm->label;
-		n = snprintf(p, len, "mc#%u", mci->mc_idx);
+		n = scnprintf(p, len, "mc#%u", mci->mc_idx);
 		p += n;
 		len -= n;
 		for (layer = 0; layer < mci->n_layers; layer++) {
-			n = snprintf(p, len, "%s#%u",
-				     edac_layer_name[mci->layers[layer].type],
-				     pos[layer]);
+			n = scnprintf(p, len, "%s#%u",
+				      edac_layer_name[mci->layers[layer].type],
+				      pos[layer]);
 			p += n;
 			len -= n;
 			dimm->location[layer] = pos[layer];
-
-			if (len <= 0)
-				break;
 		}
 
 		/* Link it to the csrows old API data */
@@ -1027,12 +1022,13 @@ void edac_mc_handle_error(const enum hw_event_mc_err_type type,
 			  const char *other_detail)
 {
 	struct dimm_info *dimm;
-	char *p;
+	char *p, *end;
 	int row = -1, chan = -1;
 	int pos[EDAC_MAX_LAYERS] = { top_layer, mid_layer, low_layer };
 	int i, n_labels = 0;
 	struct edac_raw_error_desc *e = &mci->error_desc;
 	bool any_memory = true;
+	const char *prefix;
 
 	edac_dbg(3, "MC%d\n", mci->mc_idx);
 
@@ -1087,6 +1083,8 @@ void edac_mc_handle_error(const enum hw_event_mc_err_type type,
 	 */
 	p = e->label;
 	*p = '\0';
+	end = p + sizeof(e->label);
+	prefix = "";
 
 	mci_for_each_dimm(mci, dimm) {
 		if (top_layer >= 0 && top_layer != dimm->location[0])
@@ -1114,12 +1112,8 @@ void edac_mc_handle_error(const enum hw_event_mc_err_type type,
 			p = e->label;
 			*p = '\0';
 		} else {
-			if (p != e->label) {
-				strcpy(p, OTHER_LABEL);
-				p += strlen(OTHER_LABEL);
-			}
-			strcpy(p, dimm->label);
-			p += strlen(p);
+			p += scnprintf(p, end - p, "%s%s", prefix, dimm->label);
+			prefix = OTHER_LABEL;
 		}
 
 		/*
@@ -1141,25 +1135,25 @@ void edac_mc_handle_error(const enum hw_event_mc_err_type type,
 	}
 
 	if (any_memory)
-		strcpy(e->label, "any memory");
+		strscpy(e->label, "any memory", sizeof(e->label));
 	else if (!*e->label)
-		strcpy(e->label, "unknown memory");
+		strscpy(e->label, "unknown memory", sizeof(e->label));
 
 	edac_inc_csrow(e, row, chan);
 
 	/* Fill the RAM location data */
 	p = e->location;
+	end = p + sizeof(e->location);
+	prefix = "";
 
 	for (i = 0; i < mci->n_layers; i++) {
 		if (pos[i] < 0)
 			continue;
 
-		p += sprintf(p, "%s:%d ",
-			     edac_layer_name[mci->layers[i].type],
-			     pos[i]);
+		p += scnprintf(p, end - p, "%s%s:%d", prefix,
+			       edac_layer_name[mci->layers[i].type], pos[i]);
+		prefix = " ";
 	}
-	if (p > e->location)
-		*(p - 1) = '\0';
 
 	edac_raw_mc_handle_error(e);
 }
Signature · 1 line
2.30.2
maintainer Applied
#6 lore raw

Re: [PATCH] EDAC/mc: Prefer strscpy or scnprintf over strcpy, sprintf and snprintf

Quoted 35 lines · Robert Richter

From 01a3c62a533e71984dfff7189e247b3e848f1449 Mon Sep 17 00:00:00 2001
From: Len Baker <[email protected]>
Date: Fri, 3 Sep 2021 17:05:39 +0200
Subject: [PATCH] EDAC/mc: Prefer strscpy or scnprintf over strcpy, sprintf
and snprintf

strcpy() performs no bounds checking on the destination buffer. This could result in linear overflows beyond the end of the buffer, leading to all kinds of misbehaviors. The safe replacement is strscpy(). [1][2]

However, to simplify and clarify the code, to concatenate labels use the scnprintf() function. This way it is not necessary to check the return value of strscpy (-E2BIG if the parameter count is 0 or the src was truncated) since the scnprintf returns always the number of chars written into the buffer. This function returns always a nul-terminated string even if it needs to be truncated.

While at it, fix all other broken string generation code that wrongly interprets snprintf()'s return code or just uses sprintf(), implement that using scnprintf() here too. Drop breaks in loops around scnprintf() as it is safe now to loop. Moreover, the check is
needless: For the case when the buffer is exhausted, len never gets zero because scnprintf() takes the full buffer length as input parameter, but excludes the trailing '\0' in its return code and thus, 1 is the minimum len.

[1] https://www.kernel.org/doc/html/latest/process/deprecated.html#strcpy
[2] https://github.com/KSPP/linux/issues/88

[ rric: Replace snprintf() with scnprintf(), rework sprintf() user,
drop breaks in loops around scnprintf(), introduce 'end' pointer to reduce pointer arithmetic, use prefix pattern for e->location, adjust subject and description ]

Quoted 7 lines
---
 drivers/edac/edac_mc.c | 42 ++++++++++++++++++------------------------
 1 file changed, 18 insertions(+), 24 deletions(-)

Applied, thanks.

Signature · 4 lines
Regards/Gruss,
    Boris.

https://people.kernel.org/tglx/notes-about-netiquette
author
#7 lore raw

Re: [PATCH] EDAC/mc: Prefer strscpy or scnprintf over strcpy, sprintf and snprintf

In reply to Robert Richter

Hi,

Quoted 4 lines · Robert Richter

this patch looks good to me. I made some changes on top of it to further ease pointer arithmetic and also fix remaining sprintf/snprintf() users as it makes sense to have them all in a single change. See below. Boris, please apply.

Thanks Robert for doing this.

Regards,
Len