Skip to contentSkip to conversation

add Packet hub driver for Topcliff Platform controller hub

DISCUSSION Dan Carpenter 3 participants 6 messages Jan 7, 2013 – Jan 8, 2013

lore
Share & tools
lore
#0 lore raw

Hi Masayuki Ohtak,

The patch cf4ece53460c: "add Packet hub driver for Topcliff Platform controller hub" from Sep 1, 2010, leads to the following warning:
drivers/misc/pch_phub.c:596 pch_phub_bin_write()

	 error: buffer overflow 'buf' 4096 <= 15359

Sorry my question is about an old patch. Smatch complains because we only pass a PAGE_SIZE buffer to sysfs files so the test for "if (count > PCH_PHUB_OROM_SIZE) {" makes it think we are overflowing. In fact, count is never more than 4096 so there is no overflow, but I also think that it means only the first 4096 bytes of the firmware gets updated.

drivers/misc/pch_phub.c
560 static ssize_t pch_phub_bin_write(struct file *filp, struct kobject *kobj,
561 struct bin_attribute *attr,
562 char *buf, loff_t off, size_t count)
563 {

   564          int err;
   565          unsigned int addr_offset;
   566          int ret;
   567          ssize_t rom_size;
   568          struct pch_phub_reg *chip =
   569                  dev_get_drvdata(container_of(kobj, struct device, kobj));
   570  
   571          ret = mutex_lock_interruptible(&pch_phub_mutex);
   572          if (ret)
   573                  return -ERESTARTSYS;
   574  
   575          if (off > PCH_PHUB_OROM_SIZE) {
   576                  addr_offset = 0;
   577                  goto return_ok;
   578          }
   579          if (count > PCH_PHUB_OROM_SIZE) {
                            ^^^^^^^^^^^^^^^^^^

This is 15359.

580 addr_offset = 0;
581 goto return_ok;
582 }
583
584 chip->pch_phub_extrom_base_address = pci_map_rom(chip->pdev, &rom_size);
585 if (!chip->pch_phub_extrom_base_address) {
586 err = -ENOMEM;
587 goto exrom_map_err;
588 }
589
590 for (addr_offset = 0; addr_offset < count; addr_offset++) {
591 if (PCH_PHUB_OROM_SIZE < off + addr_offset)
592 goto return_ok;
593
594 ret = pch_phub_write_serial_rom(chip,
595 chip->pch_opt_rom_start_address + addr_offset + off,

   596                              buf[addr_offset]);
                                    ^^^^^^^^^^^^^^^^

Smatch complains because "buf" is only 4096 bytes.

regards,
dan carpenter

#1 lore raw

Hi Dan,

Quoted 54 lines · Dan Carpenter

The patch cf4ece53460c: "add Packet hub driver for Topcliff Platform controller hub" from Sep 1, 2010, leads to the following warning:
drivers/misc/pch_phub.c:596 pch_phub_bin_write()

         error: buffer overflow 'buf' 4096 <= 15359

Sorry my question is about an old patch. Smatch complains because we only pass a PAGE_SIZE buffer to sysfs files so the test for "if (count > PCH_PHUB_OROM_SIZE) {" makes it think we are overflowing. In fact, count is never more than 4096 so there is no overflow, but I also think that it means only the first 4096 bytes of the firmware gets updated.

drivers/misc/pch_phub.c
560 static ssize_t pch_phub_bin_write(struct file *filp, struct kobject *kobj,
561 struct bin_attribute *attr,
562 char *buf, loff_t off, size_t count)

   563  {
   564          int err;
   565          unsigned int addr_offset;
   566          int ret;
   567          ssize_t rom_size;
   568          struct pch_phub_reg *chip =
   569                  dev_get_drvdata(container_of(kobj, struct device, kobj));
   570
   571          ret = mutex_lock_interruptible(&pch_phub_mutex);
   572          if (ret)
   573                  return -ERESTARTSYS;
   574
   575          if (off > PCH_PHUB_OROM_SIZE) {
   576                  addr_offset = 0;
   577                  goto return_ok;
   578          }
   579          if (count > PCH_PHUB_OROM_SIZE) {
                            ^^^^^^^^^^^^^^^^^^

This is 15359.

   580                  addr_offset = 0;
   581                  goto return_ok;
   582          }
   583
   584          chip->pch_phub_extrom_base_address = pci_map_rom(chip->pdev, &rom_size);
   585          if (!chip->pch_phub_extrom_base_address) {
   586                  err = -ENOMEM;
   587                  goto exrom_map_err;
   588          }
   589
   590          for (addr_offset = 0; addr_offset < count; addr_offset++) {
   591                  if (PCH_PHUB_OROM_SIZE < off + addr_offset)
   592                          goto return_ok;
   593
   594                  ret = pch_phub_write_serial_rom(chip,
   595                              chip->pch_opt_rom_start_address + addr_offset + off,
   596                              buf[addr_offset]);
                                    ^^^^^^^^^^^^^^^^
Quoted 2 lines

Smatch complains because "buf" is only 4096 bytes.

I can understand your saying.

You mean just delete the following condition ?

579 if (count > PCH_PHUB_OROM_SIZE) {

Thanks.

Signature · 2 lines
ROHM Co., Ltd.
tomoya
#3 lore raw
In reply to Dan Carpenter
Quoted 6 lines · Dan Carpenter

Sorry my question is about an old patch. Smatch complains because we only pass a PAGE_SIZE buffer to sysfs files so the test for "if (count > PCH_PHUB_OROM_SIZE) {" makes it think we are overflowing. In fact, count is never more than 4096 so there is no overflow, but I also think that it means only the first 4096 bytes of the firmware gets updated.

I think it works correctly as long as user space keeps writing into the bin file when getting short write return (4096 bytes).

Arnd

#2 lore raw
In reply to Tomoya MORINAGA
Quoted 60 lines · Tomoya MORINAGA

Hi Dan,

Quoted 54 lines · Dan Carpenter

The patch cf4ece53460c: "add Packet hub driver for Topcliff Platform controller hub" from Sep 1, 2010, leads to the following warning:
drivers/misc/pch_phub.c:596 pch_phub_bin_write()

         error: buffer overflow 'buf' 4096 <= 15359

Sorry my question is about an old patch. Smatch complains because we only pass a PAGE_SIZE buffer to sysfs files so the test for "if (count > PCH_PHUB_OROM_SIZE) {" makes it think we are overflowing. In fact, count is never more than 4096 so there is no overflow, but I also think that it means only the first 4096 bytes of the firmware gets updated.

drivers/misc/pch_phub.c
560 static ssize_t pch_phub_bin_write(struct file *filp, struct kobject *kobj,
561 struct bin_attribute *attr,
562 char *buf, loff_t off, size_t count)

   563  {
   564          int err;
   565          unsigned int addr_offset;
   566          int ret;
   567          ssize_t rom_size;
   568          struct pch_phub_reg *chip =
   569                  dev_get_drvdata(container_of(kobj, struct device, kobj));
   570
   571          ret = mutex_lock_interruptible(&pch_phub_mutex);
   572          if (ret)
   573                  return -ERESTARTSYS;
   574
   575          if (off > PCH_PHUB_OROM_SIZE) {
   576                  addr_offset = 0;
   577                  goto return_ok;
   578          }
   579          if (count > PCH_PHUB_OROM_SIZE) {
                            ^^^^^^^^^^^^^^^^^^

This is 15359.

   580                  addr_offset = 0;
   581                  goto return_ok;
   582          }
   583
   584          chip->pch_phub_extrom_base_address = pci_map_rom(chip->pdev, &rom_size);
   585          if (!chip->pch_phub_extrom_base_address) {
   586                  err = -ENOMEM;
   587                  goto exrom_map_err;
   588          }
   589
   590          for (addr_offset = 0; addr_offset < count; addr_offset++) {
   591                  if (PCH_PHUB_OROM_SIZE < off + addr_offset)
   592                          goto return_ok;
   593
   594                  ret = pch_phub_write_serial_rom(chip,
   595                              chip->pch_opt_rom_start_address + addr_offset + off,
   596                              buf[addr_offset]);
                                    ^^^^^^^^^^^^^^^^
Quoted 2 lines

Smatch complains because "buf" is only 4096 bytes.

Quoted 6 lines

I can understand your saying.

You mean just delete the following condition ?

579 if (count > PCH_PHUB_OROM_SIZE) {

What I'm saying is that sysfs files can only be 4096 bytes (larger on some arches with a larger PAGE_SIZE) and your firmware is larger than that. It won't work.

regards,
dan carpenter

Quoted 5 lines

Thanks.

Signature · 2 lines
ROHM Co., Ltd.
tomoya
#4 lore raw
In reply to Arnd Bergmann
Quoted 11 lines · Arnd Bergmann
Quoted 6 lines · Dan Carpenter

Sorry my question is about an old patch. Smatch complains because we only pass a PAGE_SIZE buffer to sysfs files so the test for "if (count > PCH_PHUB_OROM_SIZE) {" makes it think we are overflowing. In fact, count is never more than 4096 so there is no overflow, but I also think that it means only the first 4096 bytes of the firmware gets updated.

I think it works correctly as long as user space keeps writing into the bin file when getting short write return (4096 bytes).

Ah. Ok. I misunderstood. I've looked through it again and you're right.

Sorry for the noise.

regards,
dan carpenter

#5 lore raw
Quoted 9 lines · Dan Carpenter
Quoted 8 lines · Arnd Bergmann
Quoted 6 lines · Dan Carpenter

Sorry my question is about an old patch. Smatch complains because we only pass a PAGE_SIZE buffer to sysfs files so the test for "if (count > PCH_PHUB_OROM_SIZE) {" makes it think we are overflowing. In fact, count is never more than 4096 so there is no overflow, but I also think that it means only the first 4096 bytes of the firmware gets updated.

Quoted 8 lines
Quoted 3 lines

I think it works correctly as long as user space keeps writing into the bin file when getting short write return (4096 bytes).

Ah. Ok. I misunderstood. I've looked through it again and you're right.

Sorry for the noise.

No worries, the report looked totally plausible and I also had to look up how bin files work.

Arnd