summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2017-10-20 01:07:22 +0000
committerJohannes Weiner <hannes@cmpxchg.org>2017-10-20 01:07:22 +0000
commit8c953f23aaffa1931eb463adbe10f0303ef977b1 (patch)
tree63c143bc005bc8bca8812a223fe0db771226793f
parentc59626dbd4228dcf47edf7983dcb4ef21169ef95 (diff)
pci: test for unexpectedly disabled bridgesv4.14-rc5-mmotm-2017-10-19-17-00
The all-ones value is not just a "device didn't exist" case, it's also potentially a quite valid value, so not restoring it would be wrong. What *would* be interesting is to hear where the bad values came from in the first place. It sounds like the device state is saved after the PCI bus controller in front of the device has been crapped on, resulting in the PCI config cycles never reaching the device at all. Something along this patch (together with suspend/resume debugging output) migth help pinpoint it. But it really sounds like something totally brokenly turned off the PCI bridge (some ACPI shutdown crud? I wouldn't be entirely surprised) Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
-rw-r--r--drivers/pci/pci.c9
1 files changed, 9 insertions, 0 deletions
diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
index f0d68066c726..6e46fa2b6d78 100644
--- a/drivers/pci/pci.c
+++ b/drivers/pci/pci.c
@@ -1094,6 +1094,15 @@ static void pci_restore_pcix_state(struct pci_dev *dev)
int pci_save_state(struct pci_dev *dev)
{
int i;
+ u32 val;
+
+ /* Unable to read PCI device/manufacturer state? Something is seriously wrong! */
+ if (pci_read_config_dword(dev, 0, &val) || val == 0xffffffff) {
+ printk("Broken read from PCI device %s\n", pci_name(dev));
+ WARN_ON(1);
+ return -1;
+ }
+
/* XXX: 100% dword access ok here? */
for (i = 0; i < 16; i++)
pci_read_config_dword(dev, i * 4, &dev->saved_config_space[i]);