| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <CAOZ2QJPZc1tqfuQW_A6V6qTFAjJFNjXZPXaZP+NrPVmnNyoodg@mail.gmail.com>
Date: Tue, 27 Oct 2015 13:28:55 -0400
From: Dan Streetman <dan.streetman@...onical.com>
To: "Skidmore, Donald C" <donald.c.skidmore@...el.com>
Cc: "Kirsher, Jeffrey T" <jeffrey.t.kirsher@...el.com>,
"Brandeburg, Jesse" <jesse.brandeburg@...el.com>,
"Nelson, Shannon" <shannon.nelson@...el.com>,
"Wyborny, Carolyn" <carolyn.wyborny@...el.com>,
"Vick, Matthew" <matthew.vick@...el.com>,
"Ronciak, John" <john.ronciak@...el.com>,
"Williams, Mitch A" <mitch.a.williams@...el.com>,
"intel-wired-lan@...ts.osuosl.org" <intel-wired-lan@...ts.osuosl.org>,
"netdev@...r.kernel.org" <netdev@...r.kernel.org>,
"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
Dan Streetman <ddstreet@...e.org>
Subject: Re: [PATCH] ixgbe: check Master Disable bit after setting
On Tue, Oct 27, 2015 at 1:14 PM, Skidmore, Donald C
<donald.c.skidmore@...el.com> wrote:
>
>
>> -----Original Message-----
>> From: dan.streetman@...onical.com
>> [mailto:dan.streetman@...onical.com]
>> Sent: Monday, October 26, 2015 5:20 PM
>> To: Kirsher, Jeffrey T
>> Cc: Brandeburg, Jesse; Nelson, Shannon; Wyborny, Carolyn; Skidmore,
>> Donald C; Vick, Matthew; Ronciak, John; Williams, Mitch A; intel-wired-
>> lan@...ts.osuosl.org; netdev@...r.kernel.org; linux-kernel@...r.kernel.org;
>> Dan Streetman; Dan Streetman
>> Subject: [PATCH] ixgbe: check Master Disable bit after setting
>>
>> From: Dan Streetman <dan.streetman@...onical.com>
>>
>> Spec section 8.2.4.1.1 notes that after setting the PCIe Master Disable bit, it
>> must be read to verify it was set before polling the Master Enable status bit.
>>
>> This adds the check to verify the Master Disable bit was set.
>>
>> This also corrects the spec section number reference - the Master Disable
>> section is 5.2.4.3.2, not 5.2.5.3.2.
>>
>> Signed-off-by: Dan Streetman <dan.streetman@...onical.com>
>> Signed-off-by: Dan Streetman <ddstreet@...e.org>
>> ---
>> drivers/net/ethernet/intel/ixgbe/ixgbe_common.c | 13 ++++++++++++-
>> 1 file changed, 12 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_common.c
>> b/drivers/net/ethernet/intel/ixgbe/ixgbe_common.c
>> index 3f56a80..abfada7 100644
>> --- a/drivers/net/ethernet/intel/ixgbe/ixgbe_common.c
>> +++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_common.c
>> @@ -2453,6 +2453,16 @@ static s32 ixgbe_disable_pcie_master(struct
>> ixgbe_hw *hw)
>> /* Always set this bit to ensure any future transactions are blocked */
>> IXGBE_WRITE_REG(hw, IXGBE_CTRL, IXGBE_CTRL_GIO_DIS);
>>
>> + /* Spec sec 8.2.4.1.1, Master Disable bit :
>> + * "After doing any change to this bit the host must read that
>> + * the bit has been modified as expected before reading
>> + * STATUS.PCIe Master Enable Status bit."
>> + */
>> + if (!(IXGBE_READ_REG(hw, IXGBE_CTRL) & IXGBE_CTRL_GIO_DIS)) {
>> + hw_err(hw, "GIO Master Disable bit didn't set\n");
>> + goto gio_dis_fail;
>> + }
>> +
>> /* Exit if master requests are blocked */
>> if (!(IXGBE_READ_REG(hw, IXGBE_STATUS) & IXGBE_STATUS_GIO) ||
>> ixgbe_removed(hw->hw_addr))
>> @@ -2467,13 +2477,14 @@ static s32 ixgbe_disable_pcie_master(struct
>> ixgbe_hw *hw)
>>
>> /*
>> * Two consecutive resets are required via CTRL.RST per datasheet
>> - * 5.2.5.3.2 Master Disable. We set a flag to inform the reset routine
>> + * 5.2.4.3.2 Master Disable. We set a flag to inform the reset
>> +routine
>> * of this need. The first reset prevents new master requests from
>> * being issued by our device. We then must wait 1usec or more for
>> any
>> * remaining completions from the PCIe bus to trickle in, and then
>> reset
>> * again to clear out any effects they may have had on our device.
>> */
>> hw_dbg(hw, "GIO Master Disable bit didn't clear - requesting
>> resets\n");
>> +gio_dis_fail:
>> hw->mac.flags |= IXGBE_FLAGS_DOUBLE_RESET_REQUIRED;
>>
>> /*
>> --
>> 2.5.0
>
> Is this patch correcting some issue you're running in to?
I had a report of the hw hanging:
http://lists.osuosl.org/pipermail/intel-wired-lan/Week-of-Mon-20150928/002081.html
However I have no idea if this or the other patch are related at all,
I just reviewed the spec and driver and noticed these two differences;
basically a shot in the dark. The person I got the report of hw hang
from has changed their config and can't reproduce the problem anymore
- although before the change, they reproduced it several times (over a
week or so).
Do you have any ideas on how the hw would get into that state and/or
how to recover? I can send more data from the logs if it would help.
> I ask as I don't believe this check is necessary, since right below setting CTLR. PCIE_MASTER_DISABLE we loop waiting for STATUS. PCIE_MASTER_ENABLE_STATUS to clear. This should have the same effect as verify that the write to CTLR. PCIE_MASTER_DISABLE has cleared.
Well, this patch doesn't do the same check - it tries to verify that
master disable was actually set, and errors out if it wasn't, instead
of continuing on to poll enable status (which will not change, since
master disable wasn't set). That's what the spec says the driver
should do, at least.
>
> Thanks,
> Don Skidmore <donald.c.skidmore@...el.com>
>
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists