* [PATCH] ipmi: Fix kernel panic at ipmi_ssif_thread()
@ 2017-03-23 7:07 Joeseph Chang
2017-03-25 2:53 ` Corey Minyard
0 siblings, 1 reply; 5+ messages in thread
From: Joeseph Chang @ 2017-03-23 7:07 UTC (permalink / raw)
To: minyard, openipmi-developer, linux-kernel, joechang
Cc: anjiandi, Joeseph Chang
From: Joeseph Chang <joechang@codeaurora.org>
msg_written_handler() may set ssif_info->multi_data to NULL
when using ipmitool to write fru.
Change the ssif i2c send data sequence in msg_written_handler()
to fix NULL pointer kernel panic and incorrect ssif_info->multi_pos.
Signed-off-by: Joeseph Chang <joechang@codeaurora.org>
---
drivers/char/ipmi/ipmi_ssif.c | 20 +++++++++++---------
1 file changed, 11 insertions(+), 9 deletions(-)
diff --git a/drivers/char/ipmi/ipmi_ssif.c b/drivers/char/ipmi/ipmi_ssif.c
index cca6e5b..39346ee 100644
--- a/drivers/char/ipmi/ipmi_ssif.c
+++ b/drivers/char/ipmi/ipmi_ssif.c
@@ -899,21 +899,13 @@ static void msg_written_handler(struct ssif_info *ssif_info, int result,
left = 32;
/* Length byte. */
ssif_info->multi_data[ssif_info->multi_pos] = left;
- ssif_info->multi_pos += left;
- if (left < 32)
- /*
- * Write is finished. Note that we must end
- * with a write of less than 32 bytes to
- * complete the transaction, even if it is
- * zero bytes.
- */
- ssif_info->multi_data = NULL;
rv = ssif_i2c_send(ssif_info, msg_written_handler,
I2C_SMBUS_WRITE,
SSIF_IPMI_MULTI_PART_REQUEST_MIDDLE,
ssif_info->multi_data + ssif_info->multi_pos,
I2C_SMBUS_BLOCK_DATA);
+
if (rv < 0) {
/* request failed, just return the error. */
ssif_inc_stat(ssif_info, send_errors);
@@ -922,6 +914,16 @@ static void msg_written_handler(struct ssif_info *ssif_info, int result,
pr_info("Error from i2c_non_blocking_op(3)\n");
msg_done_handler(ssif_info, -EIO, NULL, 0);
}
+
+ ssif_info->multi_pos += left;
+ if (left < 32)
+ /*
+ * Write is finished. Note that we must end
+ * with a write of less than 32 bytes to
+ * complete the transaction, even if it is
+ * zero bytes.
+ */
+ ssif_info->multi_data = NULL;
} else {
/* Ready to request the result. */
unsigned long oflags, *flags;
--
1.9.1
^ permalink raw reply related [flat|nested] 5+ messages in thread* Re: [PATCH] ipmi: Fix kernel panic at ipmi_ssif_thread()
2017-03-23 7:07 [PATCH] ipmi: Fix kernel panic at ipmi_ssif_thread() Joeseph Chang
@ 2017-03-25 2:53 ` Corey Minyard
2017-03-27 3:02 ` Joseph Chang
2017-03-27 5:04 ` joechang
0 siblings, 2 replies; 5+ messages in thread
From: Corey Minyard @ 2017-03-25 2:53 UTC (permalink / raw)
To: Joeseph Chang, openipmi-developer, linux-kernel; +Cc: anjiandi, Joeseph Chang
This is incorrect. These values *must* be set before ssif_i2c_send() is
done.
Once ssif_i2c_send() is called, msg_written_handler can be called at any
time after that, including before where you moved the new code.
If I understand this correctly, I think you need to add a variable,
maybe named "bytes_to_write" and do
bytes_to_write = ssif_info->multi_data + ssif_info->multi_pos;
before the comparison and pass bytes_to_write into
ssif_i2c_send().
Thanks,
-corey
On 03/23/2017 02:07 AM, Joeseph Chang wrote:
> From: Joeseph Chang <joechang@codeaurora.org>
>
> msg_written_handler() may set ssif_info->multi_data to NULL
> when using ipmitool to write fru.
> Change the ssif i2c send data sequence in msg_written_handler()
> to fix NULL pointer kernel panic and incorrect ssif_info->multi_pos.
>
> Signed-off-by: Joeseph Chang <joechang@codeaurora.org>
> ---
> drivers/char/ipmi/ipmi_ssif.c | 20 +++++++++++---------
> 1 file changed, 11 insertions(+), 9 deletions(-)
>
> diff --git a/drivers/char/ipmi/ipmi_ssif.c b/drivers/char/ipmi/ipmi_ssif.c
> index cca6e5b..39346ee 100644
> --- a/drivers/char/ipmi/ipmi_ssif.c
> +++ b/drivers/char/ipmi/ipmi_ssif.c
> @@ -899,21 +899,13 @@ static void msg_written_handler(struct ssif_info *ssif_info, int result,
> left = 32;
> /* Length byte. */
> ssif_info->multi_data[ssif_info->multi_pos] = left;
> - ssif_info->multi_pos += left;
> - if (left < 32)
> - /*
> - * Write is finished. Note that we must end
> - * with a write of less than 32 bytes to
> - * complete the transaction, even if it is
> - * zero bytes.
> - */
> - ssif_info->multi_data = NULL;
>
> rv = ssif_i2c_send(ssif_info, msg_written_handler,
> I2C_SMBUS_WRITE,
> SSIF_IPMI_MULTI_PART_REQUEST_MIDDLE,
> ssif_info->multi_data + ssif_info->multi_pos,
> I2C_SMBUS_BLOCK_DATA);
> +
> if (rv < 0) {
> /* request failed, just return the error. */
> ssif_inc_stat(ssif_info, send_errors);
> @@ -922,6 +914,16 @@ static void msg_written_handler(struct ssif_info *ssif_info, int result,
> pr_info("Error from i2c_non_blocking_op(3)\n");
> msg_done_handler(ssif_info, -EIO, NULL, 0);
> }
> +
> + ssif_info->multi_pos += left;
> + if (left < 32)
> + /*
> + * Write is finished. Note that we must end
> + * with a write of less than 32 bytes to
> + * complete the transaction, even if it is
> + * zero bytes.
> + */
> + ssif_info->multi_data = NULL;
> } else {
> /* Ready to request the result. */
> unsigned long oflags, *flags;
^ permalink raw reply [flat|nested] 5+ messages in thread* RE: [PATCH] ipmi: Fix kernel panic at ipmi_ssif_thread()
2017-03-25 2:53 ` Corey Minyard
@ 2017-03-27 3:02 ` Joseph Chang
2017-03-27 5:04 ` joechang
1 sibling, 0 replies; 5+ messages in thread
From: Joseph Chang @ 2017-03-27 3:02 UTC (permalink / raw)
To: minyard@acm.org, openipmi-developer@lists.sourceforge.net,
linux-kernel@vger.kernel.org
Cc: Jiandi An, Joeseph Chang
Hi Corey,
Thanks your feedback and suggestion. :)
I think you are right. It is better to allocate one new local variable to store i2c data and size to be send...
I create new patch in following according to your suggestion.
Could you help to review the patch again?
>From eec31f51acc7e019fd0dcc512febda86870268d2 Mon Sep 17 00:00:00 2001
From: Joeseph Chang <joechang@codeaurora.org>
Date: Mon, 27 Mar 2017 10:46:42 +0800
Subject: [PATCH] ipmi: Fix kernel panic at ipmi_ssif_thread()
msg_written_handler() may set ssif_info->multi_data to NULL
when using ipmitool to write fru.
Fix kernel panic at ipmi_ssif_thread() and incorrect
ssif_info->multi_pos by adding one local pointer i2c_data to store i2c
data and size to send before calling ssif_i2c_send().
With this change, can avoid concurrent access to ssif_info->multi_pos and
ssif_info->multi_data at msg_written_handler.
Because msg_written_handler can be called at any time after ssif_i2c_send().
---
drivers/char/ipmi/ipmi_ssif.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)
mode change 100644 => 100755 drivers/char/ipmi/ipmi_ssif.c
diff --git a/drivers/char/ipmi/ipmi_ssif.c b/drivers/char/ipmi/ipmi_ssif.c
old mode 100644
new mode 100755
index cca6e5b..f36f018
--- a/drivers/char/ipmi/ipmi_ssif.c
+++ b/drivers/char/ipmi/ipmi_ssif.c
@@ -852,6 +852,7 @@ static void msg_written_handler(struct ssif_info *ssif_info, int result,
unsigned char *data, unsigned int len)
{
int rv;
+ unsigned char *i2c_data;
/* We are single-threaded here, so no need for a lock. */
if (result < 0) {
@@ -899,6 +900,7 @@ static void msg_written_handler(struct ssif_info *ssif_info, int result,
left = 32;
/* Length byte. */
ssif_info->multi_data[ssif_info->multi_pos] = left;
+ i2c_data = ssif_info->multi_data + ssif_info->multi_pos;
ssif_info->multi_pos += left;
if (left < 32)
/*
@@ -912,7 +914,7 @@ static void msg_written_handler(struct ssif_info *ssif_info, int result,
rv = ssif_i2c_send(ssif_info, msg_written_handler,
I2C_SMBUS_WRITE,
SSIF_IPMI_MULTI_PART_REQUEST_MIDDLE,
- ssif_info->multi_data + ssif_info->multi_pos,
+ i2c_data,
I2C_SMBUS_BLOCK_DATA);
if (rv < 0) {
/* request failed, just return the error. */
--
1.9.1
Thank you,
Joseph.
-----Original Message-----
From: Corey Minyard [mailto:tcminyard@gmail.com] On Behalf Of Corey Minyard
Sent: Saturday, March 25, 2017 10:53 AM
To: Joseph Chang <joechang@qti.qualcomm.com>; openipmi-developer@lists.sourceforge.net; linux-kernel@vger.kernel.org
Cc: Jiandi An <anjiandi@qti.qualcomm.com>; Joeseph Chang <joechang@codeaurora.org>
Subject: Re: [PATCH] ipmi: Fix kernel panic at ipmi_ssif_thread()
This is incorrect. These values *must* be set before ssif_i2c_send() is done.
Once ssif_i2c_send() is called, msg_written_handler can be called at any time after that, including before where you moved the new code.
If I understand this correctly, I think you need to add a variable, maybe named "bytes_to_write" and do
bytes_to_write = ssif_info->multi_data + ssif_info->multi_pos;
before the comparison and pass bytes_to_write into ssif_i2c_send().
Thanks,
-corey
On 03/23/2017 02:07 AM, Joeseph Chang wrote:
> From: Joeseph Chang <joechang@codeaurora.org>
>
> msg_written_handler() may set ssif_info->multi_data to NULL when using
> ipmitool to write fru.
> Change the ssif i2c send data sequence in msg_written_handler() to fix
> NULL pointer kernel panic and incorrect ssif_info->multi_pos.
>
> Signed-off-by: Joeseph Chang <joechang@codeaurora.org>
> ---
> drivers/char/ipmi/ipmi_ssif.c | 20 +++++++++++---------
> 1 file changed, 11 insertions(+), 9 deletions(-)
>
> diff --git a/drivers/char/ipmi/ipmi_ssif.c
> b/drivers/char/ipmi/ipmi_ssif.c index cca6e5b..39346ee 100644
> --- a/drivers/char/ipmi/ipmi_ssif.c
> +++ b/drivers/char/ipmi/ipmi_ssif.c
> @@ -899,21 +899,13 @@ static void msg_written_handler(struct ssif_info *ssif_info, int result,
> left = 32;
> /* Length byte. */
> ssif_info->multi_data[ssif_info->multi_pos] = left;
> - ssif_info->multi_pos += left;
> - if (left < 32)
> - /*
> - * Write is finished. Note that we must end
> - * with a write of less than 32 bytes to
> - * complete the transaction, even if it is
> - * zero bytes.
> - */
> - ssif_info->multi_data = NULL;
>
> rv = ssif_i2c_send(ssif_info, msg_written_handler,
> I2C_SMBUS_WRITE,
> SSIF_IPMI_MULTI_PART_REQUEST_MIDDLE,
> ssif_info->multi_data + ssif_info->multi_pos,
> I2C_SMBUS_BLOCK_DATA);
> +
> if (rv < 0) {
> /* request failed, just return the error. */
> ssif_inc_stat(ssif_info, send_errors); @@ -922,6 +914,16 @@
> static void msg_written_handler(struct ssif_info *ssif_info, int result,
> pr_info("Error from i2c_non_blocking_op(3)\n");
> msg_done_handler(ssif_info, -EIO, NULL, 0);
> }
> +
> + ssif_info->multi_pos += left;
> + if (left < 32)
> + /*
> + * Write is finished. Note that we must end
> + * with a write of less than 32 bytes to
> + * complete the transaction, even if it is
> + * zero bytes.
> + */
> + ssif_info->multi_data = NULL;
> } else {
> /* Ready to request the result. */
> unsigned long oflags, *flags;
^ permalink raw reply related [flat|nested] 5+ messages in thread* Re: [PATCH] ipmi: Fix kernel panic at ipmi_ssif_thread()
2017-03-25 2:53 ` Corey Minyard
2017-03-27 3:02 ` Joseph Chang
@ 2017-03-27 5:04 ` joechang
1 sibling, 0 replies; 5+ messages in thread
From: joechang @ 2017-03-27 5:04 UTC (permalink / raw)
To: minyard
Cc: Joeseph Chang, openipmi-developer, linux-kernel, anjiandi,
Corey Minyard
Hi Corey,
Thanks your feedback and suggestion. :)
I think you are right. It is better to allocate one new local variable
to store i2c data and size that will be sent...
I create new patch in following according to your suggestion.
Could you help to review the patch again?
From eec31f51acc7e019fd0dcc512febda86870268d2 Mon Sep 17 00:00:00 2001
From: Joeseph Chang <joechang@codeaurora.org>
Date: Mon, 27 Mar 2017 10:46:42 +0800
Subject: [PATCH] ipmi: Fix kernel panic at ipmi_ssif_thread()
msg_written_handler() may set ssif_info->multi_data to NULL
when using ipmitool to write fru.
Fix kernel panic at ipmi_ssif_thread() and incorrect
ssif_info->multi_pos by adding one local pointer i2c_data to store i2c
data and size to send before calling ssif_i2c_send().
With this change, can avoid concurrent access to ssif_info->multi_pos
and
ssif_info->multi_data at msg_written_handler.
Because msg_written_handler can be called at any time after
ssif_i2c_send().
---
drivers/char/ipmi/ipmi_ssif.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)
mode change 100644 => 100755 drivers/char/ipmi/ipmi_ssif.c
diff --git a/drivers/char/ipmi/ipmi_ssif.c
b/drivers/char/ipmi/ipmi_ssif.c
old mode 100644
new mode 100755
index cca6e5b..f36f018
--- a/drivers/char/ipmi/ipmi_ssif.c
+++ b/drivers/char/ipmi/ipmi_ssif.c
@@ -852,6 +852,7 @@ static void msg_written_handler(struct ssif_info
*ssif_info, int result,
unsigned char *data, unsigned int len)
{
int rv;
+ unsigned char *i2c_data;
/* We are single-threaded here, so no need for a lock. */
if (result < 0) {
@@ -899,6 +900,7 @@ static void msg_written_handler(struct ssif_info
*ssif_info, int result,
left = 32;
/* Length byte. */
ssif_info->multi_data[ssif_info->multi_pos] = left;
+ i2c_data = ssif_info->multi_data + ssif_info->multi_pos;
ssif_info->multi_pos += left;
if (left < 32)
/*
@@ -912,7 +914,7 @@ static void msg_written_handler(struct ssif_info
*ssif_info, int result,
rv = ssif_i2c_send(ssif_info, msg_written_handler,
I2C_SMBUS_WRITE,
SSIF_IPMI_MULTI_PART_REQUEST_MIDDLE,
- ssif_info->multi_data + ssif_info->multi_pos,
+ i2c_data,
I2C_SMBUS_BLOCK_DATA);
if (rv < 0) {
/* request failed, just return the error. */
--
1.9.1
Thank you,
Joseph.
Corey Minyard 於 2017-03-25 10:53 寫到:
> This is incorrect. These values *must* be set before ssif_i2c_send()
> is done.
> Once ssif_i2c_send() is called, msg_written_handler can be called at
> any
> time after that, including before where you moved the new code.
>
> If I understand this correctly, I think you need to add a variable,
> maybe named "bytes_to_write" and do
>
> bytes_to_write = ssif_info->multi_data + ssif_info->multi_pos;
>
> before the comparison and pass bytes_to_write into
> ssif_i2c_send().
>
> Thanks,
>
> -corey
>
> On 03/23/2017 02:07 AM, Joeseph Chang wrote:
>> From: Joeseph Chang <joechang@codeaurora.org>
>>
>> msg_written_handler() may set ssif_info->multi_data to NULL
>> when using ipmitool to write fru.
>> Change the ssif i2c send data sequence in msg_written_handler()
>> to fix NULL pointer kernel panic and incorrect ssif_info->multi_pos.
>>
>> Signed-off-by: Joeseph Chang <joechang@codeaurora.org>
>> ---
>> drivers/char/ipmi/ipmi_ssif.c | 20 +++++++++++---------
>> 1 file changed, 11 insertions(+), 9 deletions(-)
>>
>> diff --git a/drivers/char/ipmi/ipmi_ssif.c
>> b/drivers/char/ipmi/ipmi_ssif.c
>> index cca6e5b..39346ee 100644
>> --- a/drivers/char/ipmi/ipmi_ssif.c
>> +++ b/drivers/char/ipmi/ipmi_ssif.c
>> @@ -899,21 +899,13 @@ static void msg_written_handler(struct ssif_info
>> *ssif_info, int result,
>> left = 32;
>> /* Length byte. */
>> ssif_info->multi_data[ssif_info->multi_pos] = left;
>> - ssif_info->multi_pos += left;
>> - if (left < 32)
>> - /*
>> - * Write is finished. Note that we must end
>> - * with a write of less than 32 bytes to
>> - * complete the transaction, even if it is
>> - * zero bytes.
>> - */
>> - ssif_info->multi_data = NULL;
>> rv = ssif_i2c_send(ssif_info, msg_written_handler,
>> I2C_SMBUS_WRITE,
>> SSIF_IPMI_MULTI_PART_REQUEST_MIDDLE,
>> ssif_info->multi_data + ssif_info->multi_pos,
>> I2C_SMBUS_BLOCK_DATA);
>> +
>> if (rv < 0) {
>> /* request failed, just return the error. */
>> ssif_inc_stat(ssif_info, send_errors);
>> @@ -922,6 +914,16 @@ static void msg_written_handler(struct ssif_info
>> *ssif_info, int result,
>> pr_info("Error from i2c_non_blocking_op(3)\n");
>> msg_done_handler(ssif_info, -EIO, NULL, 0);
>> }
>> +
>> + ssif_info->multi_pos += left;
>> + if (left < 32)
>> + /*
>> + * Write is finished. Note that we must end
>> + * with a write of less than 32 bytes to
>> + * complete the transaction, even if it is
>> + * zero bytes.
>> + */
>> + ssif_info->multi_data = NULL;
>> } else {
>> /* Ready to request the result. */
>> unsigned long oflags, *flags;
^ permalink raw reply related [flat|nested] 5+ messages in thread
* [PATCH] ipmi: Fix kernel panic at ipmi_ssif_thread()
@ 2017-03-28 2:22 Joeseph Chang
0 siblings, 0 replies; 5+ messages in thread
From: Joeseph Chang @ 2017-03-28 2:22 UTC (permalink / raw)
To: minyard, openipmi-developer, linux-kernel, joechang; +Cc: anjiandi
Signed-off-by: Joeseph Chang <joechang@codeaurora.org>
---
ipmi: Fix kernel panic at ipmi_ssif_thread()
msg_written_handler() may set ssif_info->multi_data to NULL
when using ipmitool to write fru.
Before setting ssif_info->multi_data to NULL, add new local
pointer "data_to_send" and store correct i2c data pointer to
it to fix NULL pointer kernel panic and incorrect ssif_info->multi_pos.
drivers/char/ipmi/ipmi_ssif.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/drivers/char/ipmi/ipmi_ssif.c b/drivers/char/ipmi/ipmi_ssif.c
index cca6e5b..51ba67d 100644
--- a/drivers/char/ipmi/ipmi_ssif.c
+++ b/drivers/char/ipmi/ipmi_ssif.c
@@ -891,6 +891,7 @@ static void msg_written_handler(struct ssif_info *ssif_info, int result,
* for details on the intricacies of this.
*/
int left;
+ unsigned char *data_to_send;
ssif_inc_stat(ssif_info, sent_messages_parts);
@@ -899,6 +900,7 @@ static void msg_written_handler(struct ssif_info *ssif_info, int result,
left = 32;
/* Length byte. */
ssif_info->multi_data[ssif_info->multi_pos] = left;
+ data_to_send = ssif_info->multi_data + ssif_info->multi_pos;
ssif_info->multi_pos += left;
if (left < 32)
/*
@@ -912,7 +914,7 @@ static void msg_written_handler(struct ssif_info *ssif_info, int result,
rv = ssif_i2c_send(ssif_info, msg_written_handler,
I2C_SMBUS_WRITE,
SSIF_IPMI_MULTI_PART_REQUEST_MIDDLE,
- ssif_info->multi_data + ssif_info->multi_pos,
+ data_to_send,
I2C_SMBUS_BLOCK_DATA);
if (rv < 0) {
/* request failed, just return the error. */
--
1.9.1
^ permalink raw reply related [flat|nested] 5+ messages in thread
end of thread, other threads:[~2017-03-28 2:22 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2017-03-23 7:07 [PATCH] ipmi: Fix kernel panic at ipmi_ssif_thread() Joeseph Chang
2017-03-25 2:53 ` Corey Minyard
2017-03-27 3:02 ` Joseph Chang
2017-03-27 5:04 ` joechang
-- strict thread matches above, loose matches on Subject: below --
2017-03-28 2:22 Joeseph Chang
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox