s390/raw3270: Handle memory allocation failures in raw3270_setup_console()
PATCH v2 1/1 s390/raw3270: Handle memory allocation failures in raw3270_setup_console()
From: Yunseong Kim <[email protected]>
A null pointer is stored in a local variable after a call of the function "kzalloc" failed. This pointer was passed to a subsequent call of the function "raw3270_setup_device" where an undesirable dereference will be performed then. Thus add corresponding return value checks. The allocated each memory areas are immediately overwritten by the called function zero-initialisation be omitted by calling the "kmalloc" instead. After "ccw_device_enable_console" succeeds, set the bit raw3270 flag to RAW3270_FLAGS_CONSOLE.
--- drivers/s390/char/raw3270.c | 20 +++++++++++++++----- 1 file changed, 15 insertions(+), 5 deletions(-)
drivers/s390/char/raw3270.c +15 -5
@@ -811,18 +811,28 @@ struct raw3270 __init *raw3270_setup_console(void)if (IS_ERR (cdev))return ERR_CAST (cdev); - rp =kzalloc (sizeof (*rp),GFP_KERNEL |GFP_DMA ); - ascebc =kzalloc (256 ,GFP_KERNEL ); + rp =kmalloc (sizeof (*rp),GFP_KERNEL |GFP_DMA ); +if (!rp) +return ERR_PTR (-ENOMEM ); + ascebc =kmalloc (256 ,GFP_KERNEL ); +if (!ascebc) { +kfree (rp); +return ERR_PTR (-ENOMEM ); + } rc =raw3270_setup_device (cdev, rp, ascebc); -if (rc) +if (rc) { +kfree (ascebc); +kfree (rp);return ERR_PTR (rc); -set_bit (RAW3270_FLAGS_CONSOLE , &rp->flags); - + } rc =ccw_device_enable_console (cdev);if (rc) {ccw_device_destroy_console (cdev); +kfree (ascebc); +kfree (rp);return ERR_PTR (rc); } +set_bit (RAW3270_FLAGS_CONSOLE , &rp->flags);spin_lock_irqsave (get_ccwdev_lock (rp->cdev), flags); do {
Signature · 1 line
2.45.2
Heiko Carstens
maintainerQuoted 17 lines · On Tue, Jun 25, 2024 at 10:32:26AM +0900, [email protected] wrote:
From: Yunseong Kim <[email protected]>
A null pointer is stored in a local variable after a call of the function "kzalloc" failed. This pointer was passed to a subsequent call of the function "
raw3270_setup_device" where an undesirable dereference will be performed then. Thus add corresponding return value checks. The allocated each memory areas are immediately overwritten by the called function zero-initialisation be omitted by calling the "kmalloc" instead. After "ccw_device_enable_console" succeeds, set the bit raw3270 flag toRAW3270_FLAGS_CONSOLE.Fixes 33403dcfcdfd ("[S390] 3270 console: convert from bootmem to slab")Signed-off-by Yunseong Kim <[email protected]>--- drivers/s390/char/raw3270.c | 20 +++++++++++++++----- 1 file changed, 15 insertions(+), 5 deletions(-)
...
Quoted 7 lines
drivers/s390/char/raw3270.c +3
-2rc =raw3270_setup_device (cdev, rp, ascebc); -if (rc) +if (rc) { +kfree (ascebc); +kfree (rp);return ERR_PTR (rc); -set_bit (RAW3270_FLAGS_CONSOLE , &rp->flags);
drivers/s390/char/raw3270.c struct raw3270 __init *raw3270_setup_console(void) · line 835
Quoted 10 lines
drivers/s390/char/raw3270.c +4
-1- + } rc =ccw_device_enable_console (cdev);if (rc) {ccw_device_destroy_console (cdev); +kfree (ascebc); +kfree (rp);return ERR_PTR (rc); } +set_bit (RAW3270_FLAGS_CONSOLE , &rp->flags);
Why did you move the set_bit() call?
Hi Heiko,
Quoted 30 lines · Heiko Carstens
Quoted 17 lines · On Tue, Jun 25, 2024 at 10:32:26AM +0900, [email protected] wrote:
From: Yunseong Kim <[email protected]>
A null pointer is stored in a local variable after a call of the function "kzalloc" failed. This pointer was passed to a subsequent call of the function "
raw3270_setup_device" where an undesirable dereference will be performed then. Thus add corresponding return value checks. The allocated each memory areas are immediately overwritten by the called function zero-initialisation be omitted by calling the "kmalloc" instead. After "ccw_device_enable_console" succeeds, set the bit raw3270 flag toRAW3270_FLAGS_CONSOLE.Fixes 33403dcfcdfd ("[S390] 3270 console: convert from bootmem to slab")Signed-off-by Yunseong Kim <[email protected]>--- drivers/s390/char/raw3270.c | 20 +++++++++++++++----- 1 file changed, 15 insertions(+), 5 deletions(-)...
Quoted 11 lines
drivers/s390/char/raw3270.c +4
-3rc =raw3270_setup_device (cdev, rp, ascebc); -if (rc) +if (rc) { +kfree (ascebc); +kfree (rp);return ERR_PTR (rc); -set_bit (RAW3270_FLAGS_CONSOLE , &rp->flags); - + } rc =ccw_device_enable_console (cdev);if (rc) {
Quoted 8 lines
drivers/s390/char/raw3270.c struct raw3270 __init *raw3270_setup_console(void) · line 835
Quoted 6 lines
drivers/s390/char/raw3270.c +3
-0ccw_device_destroy_console (cdev); +kfree (ascebc); +kfree (rp);return ERR_PTR (rc); } +set_bit (RAW3270_FLAGS_CONSOLE , &rp->flags);Why did you move the
set_bit()call?
Thank you for the code review Heiko.
While writing patch version 2, I spent a lot of time thinking about this part. Previously, even if function "ccw_device_enable_console" failed, the flag was set to RAW3270_FLAGS_CONSOLE and returned.
I think it would be more appropriate to set the bit after everything succeeded, so I included and submitted this code in v2 patch.
I’d appreciate hearing your thoughts on this!
Warm regards,
Yunseong Kim
Heiko Carstens
maintainerQuoted 17 lines · Yunseong Kim
Hi Heiko,
Quoted 5 lines
Quoted 5 lines
drivers/s390/char/raw3270.c +1
-2-set_bit (RAW3270_FLAGS_CONSOLE , &rp->flags); - + } rc =ccw_device_enable_console (cdev);if (rc) {Quoted 8 lines
drivers/s390/char/raw3270.c struct raw3270 __init *raw3270_setup_console(void) · line 835
Quoted 6 lines
drivers/s390/char/raw3270.c +3
-0ccw_device_destroy_console (cdev); +kfree (ascebc); +kfree (rp);return ERR_PTR (rc); } +set_bit (RAW3270_FLAGS_CONSOLE , &rp->flags);Why did you move the
set_bit()call?Thank you for the code review Heiko.
Quoted 8 lines
While writing patch version 2, I spent a lot of time thinking about this part. Previously, even if function "
ccw_device_enable_console" failed, the flag was set toRAW3270_FLAGS_CONSOLEand returned.I think it would be more appropriate to set the bit after everything succeeded, so I included and submitted this code in v2 patch.
I’d appreciate hearing your thoughts on this!
"More appropriate" is not a technical reason. Please don't mix different things into a single patch. If the set_bit() call needs to be moved then you need to provide a technical reason for it.
Quoted 1 line
… Thus add corresponding return value checks.
I suggest to move this sentence into a subsequent text line.
Quoted 2 lines
The allocated each memory areas are immediately overwritten by the called function zero-initialisation be omitted by calling the "kmalloc" instead.
It seems that you stumbled on wording difficulties according to my previous patch review suggestion.
I find the intended change more appropriate for another update step.
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/process/submitting-patches.rst?h=v6.10-rc5#n168
Quoted 2 lines
After "
ccw_device_enable_console" succeeds, set the bit raw3270 flag toRAW3270_FLAGS_CONSOLE.
Why do you find such an adjustment relevant here?
Quoted 2 lines
Fixes 33403dcfcdfd ("[S390] 3270 console: convert from bootmem to slab")
Would you like to specify a “stable tag”?
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/process/stable-kernel-rules.rst?h=v6.10-rc5#n34
Quoted 1 line
---
I would appreciate a version description behind the marker line.
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/process/submitting-patches.rst?h=v6.10-rc5#n713
…
Quoted 11 lines
+++
b/drivers/s390/char/raw3270.cdrivers/s390/char/raw3270.c +4
-2@@ -811,18 +811,28 @@ struct raw3270 __init *raw3270_setup_console(void)if (IS_ERR (cdev))return ERR_CAST (cdev); - rp =kzalloc (sizeof (*rp),GFP_KERNEL |GFP_DMA ); - ascebc =kzalloc (256 ,GFP_KERNEL ); + rp =kmalloc (sizeof (*rp),GFP_KERNEL |GFP_DMA ); +if (!rp) +return ERR_PTR (-ENOMEM ); + ascebc =kmalloc (256 ,GFP_KERNEL );
drivers/s390/char/raw3270.c struct raw3270 __init *raw3270_setup_console(void) · line 826
Quoted 10 lines
drivers/s390/char/raw3270.c +7
-1+if (!ascebc) { +kfree (rp); +return ERR_PTR (-ENOMEM ); + } rc =raw3270_setup_device (cdev, rp, ascebc); -if (rc) +if (rc) { +kfree (ascebc); +kfree (rp);return ERR_PTR (rc);
…
Please take further software design options better into account.
B) scope-based resource management
https://elixir.bootlin.com/linux/v6.10-rc5/source/include/linux/slab.h#L282
Regards,
Markus