projects / thirdparty / kernel / stable-queue.git / blame
| Commit | Line | Data |
|---|---|---|
| f71f4658 GKH |
1 | From 1e1f9ca546556e508d021545861f6b5fc75a95fe Mon Sep 17 00:00:00 2001 |
| 2 | From: Brian King <brking@linux.vnet.ibm.com> | |
| 3 | Date: Fri, 17 Nov 2017 11:05:46 -0600 | |
| 4 | Subject: igbvf: Use smp_rmb rather than read_barrier_depends | |
| 5 | ||
| 6 | From: Brian King <brking@linux.vnet.ibm.com> | |
| 7 | ||
| 8 | commit 1e1f9ca546556e508d021545861f6b5fc75a95fe upstream. | |
| 9 | ||
| 10 | The original issue being fixed in this patch was seen with the ixgbe | |
| 11 | driver, but the same issue exists with igbvf as well, as the code is | |
| 12 | very similar. read_barrier_depends is not sufficient to ensure | |
| 13 | loads following it are not speculatively loaded out of order | |
| 14 | by the CPU, which can result in stale data being loaded, causing | |
| 15 | potential system crashes. | |
| 16 | ||
| 17 | Signed-off-by: Brian King <brking@linux.vnet.ibm.com> | |
| 18 | Acked-by: Jesse Brandeburg <jesse.brandeburg@intel.com> | |
| 19 | Tested-by: Aaron Brown <aaron.f.brown@intel.com> | |
| 20 | Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com> | |
| 21 | Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> | |
| 22 | ||
| 23 | --- | |
| 24 | drivers/net/ethernet/intel/igbvf/netdev.c | 2 +- | |
| 25 | 1 file changed, 1 insertion(+), 1 deletion(-) | |
| 26 | ||
| 27 | --- a/drivers/net/ethernet/intel/igbvf/netdev.c | |
| 28 | +++ b/drivers/net/ethernet/intel/igbvf/netdev.c | |
| 29 | @@ -808,7 +808,7 @@ static bool igbvf_clean_tx_irq(struct ig | |
| 30 | break; | |
| 31 | ||
| 32 | /* prevent any other reads prior to eop_desc */ | |
| 33 | - read_barrier_depends(); | |
| 34 | + smp_rmb(); | |
| 35 | ||
| 36 | /* if DD is not set pending work has not been completed */ | |
| 37 | if (!(eop_desc->wb.status & cpu_to_le32(E1000_TXD_STAT_DD))) |