diff mbox series

[v2,01/13] core/flash: Allow OPAL flash calls for size equal to flash

Message ID 1572874189-32264-2-git-send-email-debmc@linux.ibm.com
State Superseded
Headers show
Series ipmi-hiomap: Enablement for Async opal_flash_op's | expand

Checks

Context Check Description
snowpatch_ozlabs/apply_patch success Successfully applied on branch master (d75e82dbfbb9443efeb3f9a5921ac23605aab469)
snowpatch_ozlabs/snowpatch_job_snowpatch-skiboot success Test snowpatch/job/snowpatch-skiboot on branch master
snowpatch_ozlabs/snowpatch_job_snowpatch-skiboot-dco success Signed-off-by present

Commit Message

Deb McLemore Nov. 4, 2019, 1:29 p.m. UTC
From: Cyril Bur <cyril.bur@au1.ibm.com>

A bounds checking mistake prevents opal_flash_{read,write,erase} calls
from having a length equal to the size of the flash. This bug has been
present since the beginning (e7d1f60e core/flash: Add flash API) of
these calls.

Until before d6a5b53f libflash/blocklevel: Add blocklevel_smart_erase()
6/4/2017 none of our tools would have performed a single command for the
full size of the flash. It would still have been possible to persuade
`dd` to do this by using a block size equal to the size of the flash
or other mtd related tools.

Any pflash built with blocklevel_smart_erase() will perform one call to
Linux and then Skiboot for the size of flash.

Signed-off-by: Cyril Bur <cyril.bur@au1.ibm.com>
Acked-by: Samuel Mendoza-Jonas <sam@mendozajonas.com>
Reviewed-By: Alistair Popple <alistair@popple.id.au>
Signed-off-by: Stewart Smith <stewart@linux.ibm.com>
---
 core/flash.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

Comments

Vasant Hegde Nov. 5, 2019, 10:32 a.m. UTC | #1
On 11/4/19 6:59 PM, Deb McLemore wrote:
> From: Cyril Bur <cyril.bur@au1.ibm.com>
> 
> A bounds checking mistake prevents opal_flash_{read,write,erase} calls
> from having a length equal to the size of the flash. This bug has been
> present since the beginning (e7d1f60e core/flash: Add flash API) of
> these calls.
> 
> Until before d6a5b53f libflash/blocklevel: Add blocklevel_smart_erase()
> 6/4/2017 none of our tools would have performed a single command for the
> full size of the flash. It would still have been possible to persuade
> `dd` to do this by using a block size equal to the size of the flash
> or other mtd related tools.
> 
> Any pflash built with blocklevel_smart_erase() will perform one call to
> Linux and then Skiboot for the size of flash.
> 
> Signed-off-by: Cyril Bur <cyril.bur@au1.ibm.com>
> Acked-by: Samuel Mendoza-Jonas <sam@mendozajonas.com>
> Reviewed-By: Alistair Popple <alistair@popple.id.au>
> Signed-off-by: Stewart Smith <stewart@linux.ibm.com>

This is real fix. IMO it should go as independent patch without waiting for 
entire series to ready.


-Vasant
diff mbox series

Patch

diff --git a/core/flash.c b/core/flash.c
index 7fbfca2..2f041db 100644
--- a/core/flash.c
+++ b/core/flash.c
@@ -10,6 +10,7 @@ 
 
 #include <skiboot.h>
 #include <cpu.h>
+#include <inttypes.h>
 #include <lock.h>
 #include <opal.h>
 #include <opal-msg.h>
@@ -436,8 +437,10 @@  static int64_t opal_flash_op(enum flash_op op, uint64_t id, uint64_t offset,
 		goto err;
 	}
 
-	if (size >= flash->size || offset >= flash->size
+	if (size > flash->size || offset >= flash->size
 			|| offset + size > flash->size) {
+		prlog(PR_DEBUG, "Requested flash op %d beyond flash size %" PRIu64 "\n",
+				op, flash->size);
 		rc = OPAL_PARAMETER;
 		goto err;
 	}