2018-09-13 12:02:32

by Martin Schwidefsky

[permalink] [raw]
Subject: [GIT PULL] s390 patches for 4.19 #3

Hi Linus,

this is the first try to do a pull request for s390 with a signed tag.
I keep my fingers crossed that the setup works as intended.

I have used the 8A8FDAE0 signing sub-key of my 26AE5DD2 master key:

pub rsa4096/26AE5DD2 2011-10-10 [SC]
uid [ultimate] Martin Schwidefsky <[email protected]>
sub rsa4096/1F2A4B67 2011-10-10 [E]
sub rsa2048/59DAAB8E 2018-09-11 [A]
sub rsa2048/8A8FDAE0 2018-09-12 [S]

So here we go:

Please pull s390 fixes for 4.19-rc4

The following changes since commit 60c1f89241d49bacf71035470684a8d7b4bb46ea:

Merge tag 'dma-mapping-4.19-2' of git://git.infradead.org/users/hch/dma-mapping (2018-09-02 20:09:36 -0700)

are available in the git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux.git tags/s390-4.19-3

for you to fetch changes up to fa108f95c6769ec15ea59b7db00454b82afc6121:

s390/zcrypt: remove VLA usage from the AP bus (2018-09-12 11:33:37 +0200)

----------------------------------------------------------------
s390 fixes for 4.19-rc4

One fix for the zcrypt driver to correctly handle incomplete
encryption/decryption operations.

A cleanup for the aqmask/apmask parsing to avoid variable
length arrays on the stack.

----------------------------------------------------------------
Ingo Franzki (1):
s390/crypto: Fix return code checking in cbc_paes_crypt()

Martin Schwidefsky (1):
s390/zcrypt: remove VLA usage from the AP bus

arch/s390/crypto/paes_s390.c | 2 +-
drivers/s390/crypto/ap_bus.c | 86 +++++++++++++++++---------------------------
2 files changed, 34 insertions(+), 54 deletions(-)

diff --git a/arch/s390/crypto/paes_s390.c b/arch/s390/crypto/paes_s390.c
index 80b2729..ab9a0eb 100644
--- a/arch/s390/crypto/paes_s390.c
+++ b/arch/s390/crypto/paes_s390.c
@@ -208,7 +208,7 @@ static int cbc_paes_crypt(struct blkcipher_desc *desc, unsigned long modifier,
walk->dst.virt.addr, walk->src.virt.addr, n);
if (k)
ret = blkcipher_walk_done(desc, walk, nbytes - k);
- if (n < k) {
+ if (k < n) {
if (__cbc_paes_set_key(ctx) != 0)
return blkcipher_walk_done(desc, walk, -EIO);
memcpy(param.key, ctx->pk.protkey, MAXPROTKEYSIZE);
diff --git a/drivers/s390/crypto/ap_bus.c b/drivers/s390/crypto/ap_bus.c
index ec891bc..f039266 100644
--- a/drivers/s390/crypto/ap_bus.c
+++ b/drivers/s390/crypto/ap_bus.c
@@ -872,8 +872,6 @@ static int hex2bitmap(const char *str, unsigned long *bitmap, int bits)
if (bits & 0x07)
return -EINVAL;

- memset(bitmap, 0, bits / 8);
-
if (str[0] == '0' && str[1] == 'x')
str++;
if (*str == 'x')
@@ -895,25 +893,23 @@ static int hex2bitmap(const char *str, unsigned long *bitmap, int bits)
}

/*
- * str2clrsetmasks() - parse bitmask argument and set the clear and
- * the set bitmap mask. A concatenation (done with ',') of these terms
- * is recognized:
+ * modify_bitmap() - parse bitmask argument and modify an existing
+ * bit mask accordingly. A concatenation (done with ',') of these
+ * terms is recognized:
* +<bitnr>[-<bitnr>] or -<bitnr>[-<bitnr>]
* <bitnr> may be any valid number (hex, decimal or octal) in the range
* 0...bits-1; the leading + or - is required. Here are some examples:
* +0-15,+32,-128,-0xFF
* -0-255,+1-16,+0x128
* +1,+2,+3,+4,-5,-7-10
- * Returns a clear and a set bitmask. Every positive value in the string
- * results in a bit set in the set mask and every negative value in the
- * string results in a bit SET in the clear mask. As a bit may be touched
- * more than once, the last 'operation' wins: +0-255,-128 = all but bit
- * 128 set in the set mask, only bit 128 set in the clear mask.
+ * Returns the new bitmap after all changes have been applied. Every
+ * positive value in the string will set a bit and every negative value
+ * in the string will clear a bit. As a bit may be touched more than once,
+ * the last 'operation' wins:
+ * +0-255,-128 = first bits 0-255 will be set, then bit 128 will be
+ * cleared again. All other bits are unmodified.
*/
-static int str2clrsetmasks(const char *str,
- unsigned long *clrmap,
- unsigned long *setmap,
- int bits)
+static int modify_bitmap(const char *str, unsigned long *bitmap, int bits)
{
int a, i, z;
char *np, sign;
@@ -922,9 +918,6 @@ static int str2clrsetmasks(const char *str,
if (bits & 0x07)
return -EINVAL;

- memset(clrmap, 0, bits / 8);
- memset(setmap, 0, bits / 8);
-
while (*str) {
sign = *str++;
if (sign != '+' && sign != '-')
@@ -940,13 +933,10 @@ static int str2clrsetmasks(const char *str,
str = np;
}
for (i = a; i <= z; i++)
- if (sign == '+') {
- set_bit_inv(i, setmap);
- clear_bit_inv(i, clrmap);
- } else {
- clear_bit_inv(i, setmap);
- set_bit_inv(i, clrmap);
- }
+ if (sign == '+')
+ set_bit_inv(i, bitmap);
+ else
+ clear_bit_inv(i, bitmap);
while (*str == ',' || *str == '\n')
str++;
}
@@ -970,44 +960,34 @@ static int process_mask_arg(const char *str,
unsigned long *bitmap, int bits,
struct mutex *lock)
{
- int i;
+ unsigned long *newmap, size;
+ int rc;

/* bits needs to be a multiple of 8 */
if (bits & 0x07)
return -EINVAL;

+ size = BITS_TO_LONGS(bits)*sizeof(unsigned long);
+ newmap = kmalloc(size, GFP_KERNEL);
+ if (!newmap)
+ return -ENOMEM;
+ if (mutex_lock_interruptible(lock)) {
+ kfree(newmap);
+ return -ERESTARTSYS;
+ }
+
if (*str == '+' || *str == '-') {
- DECLARE_BITMAP(clrm, bits);
- DECLARE_BITMAP(setm, bits);
-
- i = str2clrsetmasks(str, clrm, setm, bits);
- if (i)
- return i;
- if (mutex_lock_interruptible(lock))
- return -ERESTARTSYS;
- for (i = 0; i < bits; i++) {
- if (test_bit_inv(i, clrm))
- clear_bit_inv(i, bitmap);
- if (test_bit_inv(i, setm))
- set_bit_inv(i, bitmap);
- }
+ memcpy(newmap, bitmap, size);
+ rc = modify_bitmap(str, newmap, bits);
} else {
- DECLARE_BITMAP(setm, bits);
-
- i = hex2bitmap(str, setm, bits);
- if (i)
- return i;
- if (mutex_lock_interruptible(lock))
- return -ERESTARTSYS;
- for (i = 0; i < bits; i++)
- if (test_bit_inv(i, setm))
- set_bit_inv(i, bitmap);
- else
- clear_bit_inv(i, bitmap);
+ memset(newmap, 0, size);
+ rc = hex2bitmap(str, newmap, bits);
}
+ if (rc == 0)
+ memcpy(bitmap, newmap, size);
mutex_unlock(lock);
-
- return 0;
+ kfree(newmap);
+ return rc;
}

/*



2018-09-14 02:27:10

by Linus Torvalds

[permalink] [raw]
Subject: Re: [GIT PULL] s390 patches for 4.19 #3

On Thu, Sep 13, 2018 at 2:02 AM Martin Schwidefsky
<[email protected]> wrote:
>
> this is the first try to do a pull request for s390 with a signed tag.
> I keep my fingers crossed that the setup works as intended.
>
> I have used the 8A8FDAE0 signing sub-key of my 26AE5DD2 master key:

The signed tag looks ok, but you haven't actually pushed out your new
signing sub-key to the keyservers, so I can't actually verify it.

So a "gpg --send-key" would be a good idea.

Linus

2018-09-14 05:39:09

by Martin Schwidefsky

[permalink] [raw]
Subject: Re: [GIT PULL] s390 patches for 4.19 #3

On Thu, 13 Sep 2018 16:26:38 -1000
Linus Torvalds <[email protected]> wrote:

> On Thu, Sep 13, 2018 at 2:02 AM Martin Schwidefsky
> <[email protected]> wrote:
> >
> > this is the first try to do a pull request for s390 with a signed tag.
> > I keep my fingers crossed that the setup works as intended.
> >
> > I have used the 8A8FDAE0 signing sub-key of my 26AE5DD2 master key:
>
> The signed tag looks ok, but you haven't actually pushed out your new
> signing sub-key to the keyservers, so I can't actually verify it.
>
> So a "gpg --send-key" would be a good idea.

I did push it to keys.gnupg.net, but the sub-key is brand new and
probably takes some time to sync on the key server network.

The key is there on pgp.mit.edu this morning, it that to one you
are using or do you prefer a different one?

--
blue skies,
Martin.

"Reality continues to ruin my life." - Calvin.


2018-09-14 05:44:29

by Linus Torvalds

[permalink] [raw]
Subject: Re: [GIT PULL] s390 patches for 4.19 #3

On Thu, Sep 13, 2018 at 7:38 PM Martin Schwidefsky
<[email protected]> wrote:
>
> I did push it to keys.gnupg.net, but the sub-key is brand new and
> probably takes some time to sync on the key server network.

That's probably it.

> The key is there on pgp.mit.edu this morning, it that to one you
> are using or do you prefer a different one?

That's the one my laptop is set up to use (I'm on the road right now),
but it didn't receive when I tried it earlier. It's probably a pool of
servers, not a single one, and it just hadn't percolated to them all.
It came in fine now when I tried as a result of this reply, and so now
the signature verifies just fine.

Thanks,

Linus