[musl] [PATCH] implement recallocarray(3)

[musl] [PATCH] implement recallocarray(3)

[Ariadne Conill]

This OpenBSD extension is similar to reallocarray(3), but
zero-initializes the new memory area.

This extension is placed in _BSD_SOURCE, like
reallocarray(3).

This patch depends on reallocarray(3) already being
present, and will not compile without it.
---
 include/stdlib.h           |  1 +
 src/malloc/recallocarray.c | 33 +++++++++++++++++++++++++++++++++
 2 files changed, 34 insertions(+)
 create mode 100644 src/malloc/recallocarray.c

diff --git a/include/stdlib.h b/include/stdlib.h
index b54a051f..a0412ad4 100644
--- a/include/stdlib.h
+++ b/include/stdlib.h
@@ -146,6 +146,7 @@ int clearenv(void);
 #define WCOREDUMP(s) ((s) & 0x80)
 #define WIFCONTINUED(s) ((s) == 0xffff)
 void *reallocarray (void *, size_t, size_t);
+void *recallocarray (void *, size_t, size_t, size_t);
 #endif
 
 #ifdef _GNU_SOURCE
diff --git a/src/malloc/recallocarray.c b/src/malloc/recallocarray.c
new file mode 100644
index 00000000..c1370edc
--- /dev/null
+++ b/src/malloc/recallocarray.c
@@ -0,0 +1,33 @@
+#define _BSD_SOURCE
+#include <errno.h>
+#include <stdlib.h>
+#include <string.h>
+
+void *recallocarray(void *ptr, size_t om, size_t m, size_t n)
+{
+	void *newptr;
+	size_t old_size, new_size;
+
+	if (n && m > -1 / n) {
+		errno = ENOMEM;
+		return 0;
+	}
+	new_size = m * n;
+
+	if (n && om > -1 / n) {
+		errno = EINVAL;
+		return 0;
+	}
+	old_size = om * n;
+
+	if (new_size <= old_size) {
+		memset((char *) ptr + new_size, 0, old_size - new_size);
+	}
+
+	newptr = reallocarray(ptr, m, n);
+	if (new_size > old_size) {
+		memset((char *) ptr + old_size, 0, new_size - old_size);
+	}
+
+	return newptr;
+}
-- 
2.28.0

Re: [musl] [PATCH] implement recallocarray(3)

[Markus Wichmann]

On Sat, Aug 01, 2020 at 08:46:58AM -0600, Ariadne Conill wrote:
> +void *recallocarray(void *ptr, size_t om, size_t m, size_t n)
> +{
> +	void *newptr;
> +	size_t old_size, new_size;
> +
> +	if (n && m > -1 / n) {
> +		errno = ENOMEM;
> +		return 0;
> +	}
> +	new_size = m * n;
> +
> +	if (n && om > -1 / n) {
> +		errno = EINVAL;
> +		return 0;
> +	}
> +	old_size = om * n;
> +
> +	if (new_size <= old_size) {
> +		memset((char *) ptr + new_size, 0, old_size - new_size);
> +	}
> +
> +	newptr = reallocarray(ptr, m, n);
> +	if (new_size > old_size) {
> +		memset((char *) ptr + old_size, 0, new_size - old_size);
> +	}
> +
> +	return newptr;
> +}

Is there a reason for the call to reallocarray? The multiplication m * n
has already been tested for overflow and executed at that point. Might
as well just call realloc() there, right?

JM2C,
Markus

Re: [musl] [PATCH] implement recallocarray(3)

[Ariadne Conill]

On Saturday, 1 August 2020 09:52:28 MDT Markus Wichmann wrote:
> On Sat, Aug 01, 2020 at 08:46:58AM -0600, Ariadne Conill wrote:
> > +void *recallocarray(void *ptr, size_t om, size_t m, size_t n)
> > +{
> > +	void *newptr;
> > +	size_t old_size, new_size;
> > +
> > +	if (n && m > -1 / n) {
> > +		errno = ENOMEM;
> > +		return 0;
> > +	}
> > +	new_size = m * n;
> > +
> > +	if (n && om > -1 / n) {
> > +		errno = EINVAL;
> > +		return 0;
> > +	}
> > +	old_size = om * n;
> > +
> > +	if (new_size <= old_size) {
> > +		memset((char *) ptr + new_size, 0, old_size - new_size);
> > +	}
> > +
> > +	newptr = reallocarray(ptr, m, n);
> > +	if (new_size > old_size) {
> > +		memset((char *) ptr + old_size, 0, new_size - old_size);
> > +	}
> > +
> > +	return newptr;
> > +}
> 
> Is there a reason for the call to reallocarray? The multiplication m * n
> has already been tested for overflow and executed at that point. Might
> as well just call realloc() there, right?

Good catch.  I decided to use reallocarray() simply for clarity, but I can 
change it to a realloc() call.

Ariadne

Re: [musl] [PATCH] implement recallocarray(3)

[Rich Felker]

On Sat, Aug 01, 2020 at 10:23:35AM -0600, Ariadne Conill wrote:
> On Saturday, 1 August 2020 09:52:28 MDT Markus Wichmann wrote:
> > On Sat, Aug 01, 2020 at 08:46:58AM -0600, Ariadne Conill wrote:
> > > +void *recallocarray(void *ptr, size_t om, size_t m, size_t n)
> > > +{
> > > +	void *newptr;
> > > +	size_t old_size, new_size;
> > > +
> > > +	if (n && m > -1 / n) {
> > > +		errno = ENOMEM;
> > > +		return 0;
> > > +	}
> > > +	new_size = m * n;
> > > +
> > > +	if (n && om > -1 / n) {
> > > +		errno = EINVAL;
> > > +		return 0;
> > > +	}
> > > +	old_size = om * n;
> > > +
> > > +	if (new_size <= old_size) {
> > > +		memset((char *) ptr + new_size, 0, old_size - new_size);
> > > +	}
> > > +
> > > +	newptr = reallocarray(ptr, m, n);
> > > +	if (new_size > old_size) {
> > > +		memset((char *) ptr + old_size, 0, new_size - old_size);
> > > +	}
> > > +
> > > +	return newptr;
> > > +}
> > 
> > Is there a reason for the call to reallocarray? The multiplication m * n
> > has already been tested for overflow and executed at that point. Might
> > as well just call realloc() there, right?
> 
> Good catch.  I decided to use reallocarray() simply for clarity, but I can 
> change it to a realloc() call.

I'm also confused why the old size has to be checked at all. There's
inherently a contract that the old size be correct for the existing
allocation; if it's not, the wrong number of members will be zeroed.
Checking whether om*n overflows does not change this and does not
catch the more general case where om is wrong.

Is the EINVAL behavior from OpenBSD, and if so, do they have a
rationale for it?

Rich

Re: [musl] [PATCH] implement recallocarray(3)

[Ariadne Conill]

Hello,

On Saturday, 1 August 2020 11:46:41 MDT Rich Felker wrote:
> On Sat, Aug 01, 2020 at 10:23:35AM -0600, Ariadne Conill wrote:
> > On Saturday, 1 August 2020 09:52:28 MDT Markus Wichmann wrote:
> > > On Sat, Aug 01, 2020 at 08:46:58AM -0600, Ariadne Conill wrote:
> > > > +void *recallocarray(void *ptr, size_t om, size_t m, size_t n)
> > > > +{
> > > > +	void *newptr;
> > > > +	size_t old_size, new_size;
> > > > +
> > > > +	if (n && m > -1 / n) {
> > > > +		errno = ENOMEM;
> > > > +		return 0;
> > > > +	}
> > > > +	new_size = m * n;
> > > > +
> > > > +	if (n && om > -1 / n) {
> > > > +		errno = EINVAL;
> > > > +		return 0;
> > > > +	}
> > > > +	old_size = om * n;
> > > > +
> > > > +	if (new_size <= old_size) {
> > > > +		memset((char *) ptr + new_size, 0, old_size - new_size);
> > > > +	}
> > > > +
> > > > +	newptr = reallocarray(ptr, m, n);
> > > > +	if (new_size > old_size) {
> > > > +		memset((char *) ptr + old_size, 0, new_size - old_size);
> > > > +	}
> > > > +
> > > > +	return newptr;
> > > > +}
> > > 
> > > Is there a reason for the call to reallocarray? The multiplication m * n
> > > has already been tested for overflow and executed at that point. Might
> > > as well just call realloc() there, right?
> > 
> > Good catch.  I decided to use reallocarray() simply for clarity, but I can
> > change it to a realloc() call.
> 
> I'm also confused why the old size has to be checked at all. There's
> inherently a contract that the old size be correct for the existing
> allocation; if it's not, the wrong number of members will be zeroed.
> Checking whether om*n overflows does not change this and does not
> catch the more general case where om is wrong.
> 
> Is the EINVAL behavior from OpenBSD, and if so, do they have a
> rationale for it?

Yes, EINVAL is mentioned in the man page.

    If ptr is not NULL and multiplying oldnmemb and size results in
    integer overflow recallocarray() returns NULL and sets errno to
    EINVAL.

They do not mention any rationale for returning EINVAL here, but they do clear 
any memory before reclaiming it on a shrink, so that may be related.  We could 
remove the check, but I added the check to be consistent with OpenBSD.

Ariadne