On 2021/8/6 2:12, Bjorn Helgaas wrote:
On Wed, Aug 04, 2021 at 09:47:08PM +0800, Dongdong Liu wrote:
Add a 10-Bit Tag check in the P2PDMA code to ensure that a device with
10-Bit Tag Requester doesn't interact with a device that does not
support 10-BIT Tag Completer. Before that happens, the kernel should
emit a warning. "echo 0 > /sys/bus/pci/devices/.../10bit_tag" to
disable 10-BIT Tag Requester for PF device.
"echo 0 > /sys/bus/pci/devices/.../sriov_vf_10bit_tag_ctl" to disable
10-BIT Tag Requester for VF device.
s/10-BIT/10-Bit/ several times.
Will fix.
Add blank lines between paragraphs.
Will fix.
Signed-off-by: Dongdong Liu <liudongdong3@xxxxxxxxxx>
---
drivers/pci/p2pdma.c | 40 ++++++++++++++++++++++++++++++++++++++++
1 file changed, 40 insertions(+)
diff --git a/drivers/pci/p2pdma.c b/drivers/pci/p2pdma.c
index 50cdde3..948f2be 100644
--- a/drivers/pci/p2pdma.c
+++ b/drivers/pci/p2pdma.c
@@ -19,6 +19,7 @@
#include <linux/random.h>
#include <linux/seq_buf.h>
#include <linux/xarray.h>
+#include "pci.h"
enum pci_p2pdma_map_type {
PCI_P2PDMA_MAP_UNKNOWN = 0,
@@ -410,6 +411,41 @@ static unsigned long map_types_idx(struct pci_dev *client)
(client->bus->number << 8) | client->devfn;
}
+static bool check_10bit_tags_vaild(struct pci_dev *a, struct pci_dev *b,
s/vaild/valid/
Or maybe s/valid/safe/ or s/valid/supported/, since "valid" isn't
quite the right word here. We want to know whether the source is
enabled to generate 10-bit tags, and if so, whether the destination
can handle them.
"if (check_10bit_tags_valid())" does not make sense because
"check_10bit_tags_valid()" is not a question with a yes/no answer.
"10bit_tags_valid()" *might* be, because "if (10bit_tags_valid())"
makes sense. But I don't think you can start with a digit.
Or maybe you want to invert the sense, e.g.,
"10bit_tags_unsupported()", since that avoids negation at the caller:
if (10bit_tags_unsupported(a, b) ||
10bit_tags_unsupported(b, a))
map_type = PCI_P2PDMA_MAP_NOT_SUPPORTED;
Good suggestion. add a pci_ prefix.
if (pci_10bit_tags_unsupported(a, b) ||
pci_10bit_tags_unsupported(b, a))
map_type = PCI_P2PDMA_MAP_NOT_SUPPORTED;
Doesn't this patch need to be at the very beginning, before you start
enabling 10-bit tags? Otherwise there's a hole in the middle where we
enable them and P2P DMA might break.
Yes, will do.
+ bool verbose)
+{
+ bool req;
+ bool comp;
+ u16 ctl2;
+
+ if (a->is_virtfn) {
+#ifdef CONFIG_PCI_IOV
+ req = !!(a->physfn->sriov->ctrl &
+ PCI_SRIOV_CTRL_VF_10BIT_TAG_REQ_EN);
+#endif
+ } else {
+ pcie_capability_read_word(a, PCI_EXP_DEVCTL2, &ctl2);
+ req = !!(ctl2 & PCI_EXP_DEVCTL2_10BIT_TAG_REQ_EN);
+ }
+
+ comp = !!(b->pcie_devcap2 & PCI_EXP_DEVCAP2_10BIT_TAG_COMP);
+ if (req && (!comp)) {
+ if (verbose) {
+ pci_warn(a, "cannot be used for peer-to-peer DMA as 10-Bit Tag Requester enable is set in device (%s), but peer device (%s) does not support the 10-Bit Tag Completer\n",
+ pci_name(a), pci_name(b));
No point in printing pci_name(a) twice. pci_warn() prints it already;
that should be enough.
Will fix.
I think you can simplify this a little, e.g.,
if (!req) /* 10-bit tags not enabled on requester */
return true;
if (comp) /* completer can handle anything */
return true;
/* error case */
if (!verbose)
return false;
pci_warn(...);
return false;
Good point, this will make code more clean and readable.
Thanks,
Dongdong
+ if (a->is_virtfn)
+ pci_warn(a, "to disable 10-Bit Tag Requester for this device, echo 0 > /sys/bus/pci/devices/%s/sriov_vf_10bit_tag_ctl\n",
+ pci_name(a));
+ else
+ pci_warn(a, "to disable 10-Bit Tag Requester for this device, echo 0 > /sys/bus/pci/devices/%s/10bit_tag\n",
+ pci_name(a));
+ }
+ return false;
+ }
+
+ return true;
+}
+
/*
* Calculate the P2PDMA mapping type and distance between two PCI devices.
*
@@ -532,6 +568,10 @@ calc_map_type_and_dist(struct pci_dev *provider, struct pci_dev *client,
map_type = PCI_P2PDMA_MAP_NOT_SUPPORTED;
}
done:
+ if (!check_10bit_tags_vaild(client, provider, verbose) ||
+ !check_10bit_tags_vaild(provider, client, verbose))
+ map_type = PCI_P2PDMA_MAP_NOT_SUPPORTED;
+
rcu_read_lock();
p2pdma = rcu_dereference(provider->p2pdma);
if (p2pdma)
--
2.7.4
.
