2007-10-26 09:44:37

by Jeff Garzik

[permalink] [raw]
Subject: [PATCH v2] ISDN/capidrv: address two longstanding warnings

* change #warning to a code comment

* add comment and special ifdef'd 64-bit code for a situation where
we must store a pointer into a CAPI field 'Data', which is fixed by
the interface at 32 bits.

Signed-off-by: Jeff Garzik <[email protected]>
---
drivers/isdn/capi/capidrv.c | 25 +++++++++++++++++++++++--
1 files changed, 23 insertions(+), 2 deletions(-)

diff --git a/drivers/isdn/capi/capidrv.c b/drivers/isdn/capi/capidrv.c
index 476012b..48c1775 100644
--- a/drivers/isdn/capi/capidrv.c
+++ b/drivers/isdn/capi/capidrv.c
@@ -1843,6 +1843,7 @@ static int if_sendbuf(int id, int channel, int doack, struct sk_buff *skb)
int msglen;
u16 errcode;
u16 datahandle;
+ u32 data;

if (!card) {
printk(KERN_ERR "capidrv: if_sendbuf called with invalid driverId %d!\n",
@@ -1860,9 +1861,26 @@ static int if_sendbuf(int id, int channel, int doack, struct sk_buff *skb)
return 0;
}
datahandle = nccip->datahandle;
+
+ /*
+ * Here we copy pointer skb->data into the 32-bit 'Data' field.
+ * The 'Data' field is not used in practice in linux kernel
+ * (neither in 32 or 64 bit), but should have some value,
+ * since a CAPI message trace will display it.
+ *
+ * The correct value in the 32 bit case is the address of the
+ * data, in 64 bit it makes no sense, we use 0 there.
+ */
+
+#ifdef CONFIG_64BIT
+ data = 0;
+#else
+ data = (unsigned long) skb->data;
+#endif
+
capi_fill_DATA_B3_REQ(&sendcmsg, global.ap.applid, card->msgid++,
nccip->ncci, /* adr */
- (u32) skb->data, /* Data */
+ data, /* Data */
skb->len, /* DataLength */
datahandle, /* DataHandle */
0 /* Flags */
@@ -2123,7 +2141,10 @@ static int capidrv_delcontr(u16 contr)
printk(KERN_ERR "capidrv: delcontr: no contr %u\n", contr);
return -1;
}
- #warning FIXME: maybe a race condition the card should be removed here from global list /kkeil
+
+ /* FIXME: maybe a race condition the card should be removed
+ * here from global list /kkeil
+ */
spin_unlock_irqrestore(&global_lock, flags);

del_timer(&card->listentimer);
--
1.5.2.4


2007-10-26 16:44:18

by Karsten Keil

[permalink] [raw]
Subject: Re: [PATCH v2] ISDN/capidrv: address two longstanding warnings

* change #warning to a code comment

* add comment and special ifdef'd 64-bit code for a situation where
we must store a pointer into a CAPI field 'Data', which is fixed by
the interface at 32 bits.

Signed-off-by: Jeff Garzik <[email protected]>
Acked-by: Karsten Keil <[email protected]>

---
drivers/isdn/capi/capidrv.c | 25 +++++++++++++++++++++++--
1 files changed, 23 insertions(+), 2 deletions(-)

diff --git a/drivers/isdn/capi/capidrv.c b/drivers/isdn/capi/capidrv.c
index 476012b..48c1775 100644
--- a/drivers/isdn/capi/capidrv.c
+++ b/drivers/isdn/capi/capidrv.c
@@ -1843,6 +1843,7 @@ static int if_sendbuf(int id, int channel, int doack, struct sk_buff *skb)
int msglen;
u16 errcode;
u16 datahandle;
+ u32 data;

if (!card) {
printk(KERN_ERR "capidrv: if_sendbuf called with invalid driverId %d!\n",
@@ -1860,9 +1861,26 @@ static int if_sendbuf(int id, int channel, int doack, struct sk_buff *skb)
return 0;
}
datahandle = nccip->datahandle;
+
+ /*
+ * Here we copy pointer skb->data into the 32-bit 'Data' field.
+ * The 'Data' field is not used in practice in linux kernel
+ * (neither in 32 or 64 bit), but should have some value,
+ * since a CAPI message trace will display it.
+ *
+ * The correct value in the 32 bit case is the address of the
+ * data, in 64 bit it makes no sense, we use 0 there.
+ */
+
+#ifdef CONFIG_64BIT
+ data = 0;
+#else
+ data = (unsigned long) skb->data;
+#endif
+
capi_fill_DATA_B3_REQ(&sendcmsg, global.ap.applid, card->msgid++,
nccip->ncci, /* adr */
- (u32) skb->data, /* Data */
+ data, /* Data */
skb->len, /* DataLength */
datahandle, /* DataHandle */
0 /* Flags */
@@ -2123,7 +2141,10 @@ static int capidrv_delcontr(u16 contr)
printk(KERN_ERR "capidrv: delcontr: no contr %u\n", contr);
return -1;
}
- #warning FIXME: maybe a race condition the card should be removed here from global list /kkeil
+
+ /* FIXME: maybe a race condition the card should be removed
+ * here from global list /kkeil
+ */
spin_unlock_irqrestore(&global_lock, flags);

del_timer(&card->listentimer);
--
1.5.2.4

--
Karsten Keil
SuSE Labs
ISDN and VOIP development
SUSE LINUX Products GmbH, Maxfeldstr.5 90409 Nuernberg, GF: Markus Rex, HRB 16746 (AG Nuernberg)

2007-10-26 22:42:13

by Jeff Garzik

[permalink] [raw]
Subject: Re: [PATCH v2] ISDN/capidrv: address two longstanding warnings

Karsten Keil wrote:
> * change #warning to a code comment
>
> * add comment and special ifdef'd 64-bit code for a situation where
> we must store a pointer into a CAPI field 'Data', which is fixed by
> the interface at 32 bits.
>
> Signed-off-by: Jeff Garzik <[email protected]>
> Acked-by: Karsten Keil <[email protected]>

thanks!