On 7/16/2019 7:48 AM, KyleMahlkuch wrote:
Power and x86 have different page sizes so rather than allocate the
buffer based on number of pages we should allocate space by using
max_sectors. There is also code in lpfc_scsi.c to be sure we don't
write past the end of this buffer.
Signed-off-by: KyleMahlkuch <kmahl...@linux.vnet.ibm.com>
---
drivers/scsi/lpfc/lpfc_init.c | 41
+++++++----------------------------------
drivers/scsi/lpfc/lpfc_scsi.c | 14 ++++++++++++--
2 files changed, 19 insertions(+), 36 deletions(-)
Kyle,
After looking at this code, there are several issues:
- why the 2^10 (4MB @ 4k pgsz) buffer size (horribly named as pagecnt)
- didn't comprehend pagesizes > 4k
- corresponding dif buffer doesn't need to be the same size
- no checking of exceeding buffer length when copying
- allocation even when debugfs wasn't configured
- could the entity downloading identify which phba loaded the buffer ?
Given this was added a decade ago with initial DIF support and has
rarely been used, we're going to remove the debugfs bg buffer mechanism
completely.
- james
