If there is only a single bridge on the bus, we assign all resources to it.
Currently this is done as a part of the resource distribution loop but it
does not have to be there, and moving it outside actually improves
readability because we can then save one indent level in the loop.
While there we can add hotplug_bridges == 1 && normal_bridges == 0 to
the same block because they are dealt the same way.
Suggested-by: Bjorn Helgaas <bhelgaas@google.com>
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
When distributing extra bus number space to hotplug bridges for future
extension, we don't account for the fact that there might be non-hotplug
bridges on the bus after the hotplug bridges. For example:
01:00.0 --+- 02:00.0 (HotPlug-) -- Thunderbolt host controller
+- 02:01.0 (HotPlug+)
\- 02:02.0 (HotPlug-) -- xHCI host controller
pci_scan_child_bus_extend() is supposed to distribute the remaining bus
numbers to the hotplug bridge at 02:01.0, but only after accounting for all
bridges on bus 02. Since we don't check whether there's another
non-hotplug bridge after the hotplug bridge 02:01.0, it may not leave space
for the non-hotplug bridge:
pci 0000:00:1b.0: PCI bridge to [bus 01-39] (Root Port)
pci 0000:01:00.0: PCI bridge to [bus 02-39]
...
pci 0000:02:00.0: PCI bridge to [bus 03]
pci 0000:02:01.0: PCI bridge to [bus 04]
pci_bus 0000:04: [bus 04-39] extended by 0x35
pci_bus 0000:04: bus scan returning with max=39
pci_bus 0000:04: busn_res: [bus 04-39] end is updated to 39
pci 0000:02:02.0: scanning [bus 00-00] behind bridge, pass 1
pci_bus 0000:3a: scanning bus
pci_bus 0000:3a: bus scan returning with max=3a
pci_bus 0000:3a: busn_res: [bus 3a] end is updated to 3a
pci_bus 0000:3a: [bus 3a] partially hidden behind bridge 0000:02 [bus 02-39]
pci_bus 0000:3a: [bus 3a] partially hidden behind bridge 0000:01 [bus 01-39]
pci_bus 0000:02: bus scan returning with max=3a
pci_bus 0000:02: busn_res: [bus 02-39] end can not be updated to 3a
The resulting 'lspci -t' output looks like this:
+-1b.0-[01-39]----00.0-[02-3a]--+-00.0-[03]----00.0
^^ +-01.0-[04-39]--
\-02.0-[3a]----00.0
^^
The xHCI host controller behind 02:02.0 is not usable because it would have
to be assigned bus 3a, which is not accessible through 00:1b.0.
To fix this, reserve at least one bus for each bridge while scanning
already configured bridges. Then use this information in the second
scan to correct the available extra bus space for hotplug bridges.
After this change the 'lspci -t' output is what is expected:
+-1b.0-[01-39]----00.0-[02-39]--+-00.0-[03]----00.0
+-01.0-[04-38]--
\-02.0-[39]----00.0
The xHCI controller is now on bus 39, where it is usable.
Fixes: 1c02ea8100 ("PCI: Distribute available buses to hotplug-capable bridges")
Reported-by: Mario Limonciello <mario.limonciello@dell.com>
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
[bhelgaas: changelog]
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cc: stable@vger.kernel.org
Following PCIehp mark the unplugged PCI devices disconnected. This makes
sure PCI core code leaves the now missing hardware registers alone.
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
When acpiphp re-enumerates a PCI hierarchy because of an ACPI Notify()
event, we should skip bridges managed by native hotplug (pciehp or shpchp).
We don't want to scan below a native hotplug bridge until the hotplug
controller generates a hot-add event.
A typical scenario is a Root Port leading to a Thunderbolt host router that
remains powered off until something is connected to it. See [1] for the
lspci details.
1. Before something is connected, only the Root Port exists. It has
PCI_EXP_SLTCAP_HPC set and pciehp is responsible for hotplug:
00:1b.0 Root Port (HotPlug+)
2. When a USB-C or Thunderbolt device is connected, the Switch in the
Thunderbolt host router is powered up, the Root Port signals a hotplug
add event and pciehp enumerates the Switch:
01:00.0 Switch Upstream Port to [bus 02-39]
02:00.0 Switch Downstream Port to [bus 03] (HotPlug-, to NHI)
02:01.0 Switch Downstream Port to [bus 04-38] (HotPlug+, to Thunderbolt connector)
02:02.0 Switch Downstream Port to [bus 39] (HotPlug-, to xHCI)
The 02:00.0 and 02:02.0 Ports lead to Endpoints that are not powered
up yet. The Ports have PCI_EXP_SLTCAP_HPC cleared, so pciehp doesn't
handle hotplug for them and we assign minimal resources to them.
The 02:01.0 Port has PCI_EXP_SLTCAP_HPC set, so pciehp handles native
hotplug events for it.
3. The BIOS powers up the xHCI controller. If a Thunderbolt device was
connected (not just a USB-C device), it also powers up the NHI. Then
it sends an ACPI Notify() to the Root Port, and acpiphp enumerates the
new device(s):
03:00.0 Thunderbolt Host Controller (NHI) Endpoint
39:00.0 xHCI Endpoint
4. If a Thunderbolt device was connected, the host router firmware uses
the NHI to set up Thunderbolt tunnels and triggers a native hotplug
event (via 02:01.0 in this example). Then pciehp enumerates the new
Thunderbolt devices:
04:00.0 Switch Upstream Port to [bus 05-38]
05:01.0 Switch Downstream Port to [bus 06-09] (HotPlug-)
05:04.0 Switch Downstream Port to [bus 0a-38] (HotPlug+)
In this example, 05:01.0 leads to another Switch and some NICs. This
subtree is static, so 05:01.0 doesn't support hotplug and has
PCI_EXP_SLTCAP_HPC cleared.
In step 3, acpiphp previously enumerated everything below the Root Port,
including things below the 02:01.0 Port. We don't want that because pciehp
expects to manage hotplug below that Port, and firmware on the host router
may be in the middle of configuring its Link so it may not be ready yet.
To make this work better with the native PCIe (pciehp) and standard PCI
(shpchp) hotplug drivers, we let them handle all slot management and
resource allocation for hotplug bridges and restrict ACPI hotplug to
non-hotplug bridges.
[1] https://bugzilla.kernel.org/show_bug.cgi?id=199581#c5
Link: https://lkml.kernel.org/r/20180529160155.1738-1-mika.westerberg@linux.intel.com
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
[bhelgaas: changelog, use hotplug_is_native() instead of
dev->is_hotplug_bridge]
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Add hotplug_is_native() to find out whether the OS is supposed to handle
native hotplug of a given bridge.
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
In the same way we do for pciehp, add shpchp_is_native(), which returns
true if the bridge should be handled by the native SHPC driver. Then
convert the driver to use this function.
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
The fix for an AMD POGO erratum related to SHPC incorrectly identified the
device. The workaround should be applied only for AMD POGO devices, but it
was instead applied to:
- all AMD bridges, and
- all devices from any vendor with device ID 0x7458
Fixes: 53044f3574 ("[PATCH] PCI Hotplug: shpchp: AMD POGO errata fix")
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Add a driver for Mobiveil AXI PCIe Host Bridge Soft IP - GPEX 4.0,
a PCIe gen4 IP. This IP has upto 8 outbound and inbound windows
for the address translation.
Signed-off-by: Subrahmanya Lingappa <l.subrahmanya@mobiveil.co.in>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
[bhelgaas: fold in mobiveil_pcie_of_match[] NULL termination from Wei
Yongjun <weiyongjun1@huawei.com>]
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Cc: Marc Zyngier <marc.zyngier@arm.com>
Issuing "ip link set up/down" can block TSICR interrupts, what results in
missing PTP Tx timestamp and no PPS pulse generation.
Problem happens when the link is set up with the TSICR interrupts pending.
ICR is cleared before enabling interrupts, while TSICR is not. When all TSICR
interrupts are pending at this moment, time_sync interrupt will never
be generated. TSICR should be cleared as well.
In order to reproduce the issue:
1. Setup linux with IEEE 1588 grandmaster and PPS output enabled
2. Continue setting link up/down with random intervals between commands
3. Wait until PPS is not generated ( only one pulse is generated and PPS
dies), and ptp4l complains constantly about Tx timeout.
Signed-off-by: Joanna Yurdal <jyu@trackman.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
On arm64, ebt_entry_{match,watcher,target} structs are 40 bytes long
while on 32-bit arm these structs have a size of 36 bytes.
COMPAT_XT_ALIGN() macro cannot be used here to determine the necessary
padding for the CONFIG_COMPAT because it imposes an 8-byte boundary
alignment, condition that is not found in 32-bit ebtables application.
Signed-off-by: Alin Nastac <alin.nastac@gmail.com>
Acked-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Pull rmdir update from Al Viro:
"More shrink_dcache_parent()-related stuff - killing the main source of
potentially contended calls of that on large subtrees"
* 'work.rmdir' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
rmdir(),rename(): do shrink_dcache_parent() only on success
Updated the e1000.txt kernel documentation with the latest information.
Also convert the text file to reStructuredText (RST) format, since the
Linux kernel documentation now uses this format for documentation.
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
This patch avoids dereferencing msm_host->dev when it is NULL.
If we find ourselves tearing down dsi before calling
(mdp4|mdp5|dpu)_kms_init(), we'll end up in a state where the dev
pointer is NULL and trying to extract priv from it will fail.
This was introduced in a seemingly innocuous commit to ensure the
arguments to msm_gem_put_iova() are correct (even though that
function has been a stub for ~5 years). Correctness FTW! \o/
Fixes: b01884a286b0 drm/msm: use correct aspace pointer in msm_gem_put_iova()
Cc: Daniel Mack <daniel@zonque.org>
Cc: Rob Clark <robdclark@gmail.com>
Signed-off-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Rob Clark <robdclark@gmail.com>
Now that all of the msm-specific goo is tucked safely away we can switch
over to using the atomic helper commit directly. \o/
Changes in v2:
- None
Changes in v3:
- Rebased on Archit's private_obj set
Changes in v4:
- None
Cc: Abhinav Kumar <abhinavk@codeaurora.org>
Signed-off-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Rob Clark <robdclark@gmail.com>
Moving further towards switching fully to the the atomic helpers, this
patch removes the hand-rolled worker nonblock commit code and uses the
atomic helpers commit_work model.
Changes in v2:
- Remove commit_destroy()
- Shuffle order of commit_tail calls to further serialize commits
- Use stall in swap_state to avoid abandoned events on disable
Changes in v3:
- Rebased on Archit's private_obj set
Changes in v4:
- None
Signed-off-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Rob Clark <robdclark@gmail.com>
Ensure that any queued events are issued when disabling the crtc. This
avoids timeouts when we come back and wait for dependencies (like the
previous frame's flip_done).
Changes in v2:
- None
Changes in v3:
- Rebased on Archit's private_obj set
Changes in v4:
- None
Reviewed-by: Archit Taneja <architt@codeaurora.org>
Signed-off-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Rob Clark <robdclark@gmail.com>
In preparation for moving to atomic helpers, move the implicit sync
fence handling out of atomic commit and into the plane->prepare_fb()
hook. While we're at it, de-duplicate the mdp*_prepare_fb functions.
Changes in v4:
- Added
Reported-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Rob Clark <robdclark@gmail.com>
Factor out the commit_tail() portions of complete_commit() into a
separate function to facilitate moving to the atomic helpers in future
patches.
Changes in v2:
- None
Changes in v3:
- Rebased on Archit's private_obj set
Changes in v4:
- None
Cc: Jeykumar Sankaran <jsanka@codeaurora.org>
Reviewed-by: Archit Taneja <architt@codeaurora.org>
Signed-off-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Rob Clark <robdclark@gmail.com>
Use dma_zalloc_coherent for allocating zeroed memory and
remove unnecessary memset function.
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Over the years, several of the links have changed or are no longer valid
so update them. In addition, the default values were incorrect for a
couple of parameters.
Converted the text file to the reStructuredText (RST) format, since the
Linux kernel documentation now uses this format for documentation.
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
There have been multiple reports of crashes that look like
kernel: RIP: 0010:[<ffffffff8110303f>] timecounter_read+0xf/0x50
[...]
kernel: Call Trace:
kernel: [<ffffffffa0806b0f>] e1000e_phc_gettime+0x2f/0x60 [e1000e]
kernel: [<ffffffffa0806c5d>] e1000e_systim_overflow_work+0x1d/0x80 [e1000e]
kernel: [<ffffffff810992c5>] process_one_work+0x155/0x440
kernel: [<ffffffff81099e16>] worker_thread+0x116/0x4b0
kernel: [<ffffffff8109f422>] kthread+0xd2/0xf0
kernel: [<ffffffff8163184f>] ret_from_fork+0x3f/0x70
These can be traced back to the fact that e1000e_systim_reset() skips the
timecounter_init() call if e1000e_get_base_timinca() returns -EINVAL, which
leads to a null deref in timecounter_read().
Commit 83129b37ef ("e1000e: fix systim issues", v4.2-rc1) reworked
e1000e_get_base_timinca() in such a way that it can return -EINVAL for
e1000_pch_spt if the SYSCFI bit is not set in TSYNCRXCTL.
Some experimentation has shown that on I219 (e1000_pch_spt, "MAC: 12")
adapters, the E1000_TSYNCRXCTL_SYSCFI flag is unstable; TSYNCRXCTL reads
sometimes don't have the SYSCFI bit set. Retrying the read shortly after
finds the bit to be set. This was observed at boot (probe) but also link up
and link down.
Moreover, the phc (PTP Hardware Clock) seems to operate normally even after
reads where SYSCFI=0. Therefore, remove this register read and
unconditionally set the clock parameters.
Reported-by: Achim Mildenberger <admin@fph.physik.uni-karlsruhe.de>
Message-Id: <20180425065243.g5mqewg5irkwgwgv@f2>
Bugzilla: https://bugzilla.suse.com/show_bug.cgi?id=1075876
Fixes: 83129b37ef ("e1000e: fix systim issues")
Signed-off-by: Benjamin Poirier <bpoirier@suse.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
There is mistake in the rt_mode_allow_non_local assignment.
It should be used to check if sending to non-local addresses is
allowed, now it checks if local addresses are allowed.
As local addresses are allowed for most of the cases, the only
places that are affected are for traffic to transparent cache
servers:
- bypass connections when cache server is not available
- related ICMP in FORWARD hook when sent to cache server
Fixes: 4a4739d56b ("ipvs: Pull out crosses_local_route_boundary logic")
Signed-off-by: Julian Anastasov <ja@ssi.bg>
Acked-by: Simon Horman <horms@verge.net.au>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
In order to allocate icmpv6 skb, sizeof(struct ipv6hdr) should be used.
Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Again, when revalidating the inode, we don't need to ask for attributes
for which we are authoritative.
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Allow the getattr() callback to check things like whether or not we hold
a delegation so that it can adjust the attributes that it is asking for.
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
When we hold a delegation, we should not need to request attributes such
as the file size or the change attribute. For some servers, avoiding
asking for these unneeded attributes can improve the overall system
performance.
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Pull dcache updates from Al Viro:
"This is the first part of dealing with livelocks etc around
shrink_dcache_parent()."
* 'work.dcache' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
restore cond_resched() in shrink_dcache_parent()
dput(): turn into explicit while() loop
dcache: move cond_resched() into the end of __dentry_kill()
d_walk(): kill 'finish' callback
d_invalidate(): unhash immediately
Add support for "VMWRITE to any supported field in the VMCS" and
enable this feature by default in L1's IA32_VMX_MISC MSR. If userspace
clears the VMX capability bit, the old behavior will be restored.
Note that this feature is a prerequisite for kvm in L1 to use VMCS
shadowing, once that feature is available.
Signed-off-by: Jim Mattson <jmattson@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Disallow changes to the VMX capability MSRs while the vCPU is in VMX
operation. Although this does break the existing API, it helps to
avoid some potentially tricky situations for which there is no
architected behavior.
Signed-off-by: Jim Mattson <jmattson@google.com>
Reviewed-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
'Commit d0659d946b ("KVM: x86: add option to advance tscdeadline
hrtimer expiration")' advances the tscdeadline (the timer is emulated
by hrtimer) expiration in order that the latency which is incurred
by hypervisor (apic_timer_fn -> vmentry) can be avoided. This patch
adds the advance tscdeadline expiration support to which the tscdeadline
timer is emulated by VMX preemption timer to reduce the hypervisor
lantency (handle_preemption_timer -> vmentry). The guest can also
set an expiration that is very small (for example in Linux if an
hrtimer feeds a expiration in the past); in that case we set delta_tsc
to 0, leading to an immediately vmexit when delta_tsc is not bigger than
advance ns.
This patch can reduce ~63% latency (~4450 cycles to ~1660 cycles on
a haswell desktop) for kvm-unit-tests/tscdeadline_latency when testing
busy waits.
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Radim Krčmář <rkrcmar@redhat.com>
Signed-off-by: Wanpeng Li <wanpengli@tencent.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
ucma_process_join() will free the new allocated "mc" struct,
if there is any error after that, especially the copy_to_user().
But in parallel, ucma_leave_multicast() could find this "mc"
through idr_find() before ucma_process_join() frees it, since it
is already published.
So "mc" could be used in ucma_leave_multicast() after it is been
allocated and freed in ucma_process_join(), since we don't refcnt
it.
Fix this by separating "publish" from ID allocation, so that we
can get an ID first and publish it later after copy_to_user().
Fixes: c8f6a362bf ("RDMA/cma: Add multicast communication support")
Reported-by: Noam Rathaus <noamr@beyondsecurity.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Fixes: db9d7d36ee ("net: mvpp2: Split the PPv2 driver to a dedicated directory")
Signed-off-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The newly added fill_res_ep_entry function fails to link if
CONFIG_INFINIBAND_ADDR_TRANS is not set:
drivers/infiniband/hw/cxgb4/restrack.o: In function `fill_res_ep_entry':
restrack.c:(.text+0x3cc): undefined reference to `rdma_res_to_id'
restrack.c:(.text+0x3d0): undefined reference to `rdma_iw_cm_id'
This adds a Kconfig dependency for the driver.
Fixes: 116aeb8873 ("iw_cxgb4: provide detailed provider-specific CM_ID information")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Greg Thelen <gthelen@google.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Add documentation for r8a77990 compatible string to renesas ravb device
tree bindings documentation.
Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Acked-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Reviewed-by: Simon Horman <horms+renesas@verge.net.au>
Signed-off-by: Rob Herring <robh@kernel.org>
The function aq_fw2x_get_mac_permanent is local to the source and does
not need to be in global scope, so make it static.
Cleans up sparse warning:
warning: symbol 'aq_fw2x_get_mac_permanent' was not declared. Should it
be static?
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Here, the xdpsock sample application is adjusted to the new descriptor
format.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Currently, AF_XDP only supports a fixed frame-size memory scheme where
each frame is referenced via an index (idx). A user passes the frame
index to the kernel, and the kernel acts upon the data. Some NICs,
however, do not have a fixed frame-size model, instead they have a
model where a memory window is passed to the hardware and multiple
frames are filled into that window (referred to as the "type-writer"
model).
By changing the descriptor format from the current frame index
addressing scheme, AF_XDP can in the future be extended to support
these kinds of NICs.
In the index-based model, an idx refers to a frame of size
frame_size. Addressing a frame in the UMEM is done by offseting the
UMEM starting address by a global offset, idx * frame_size + offset.
Communicating via the fill- and completion-rings are done by means of
idx.
In this commit, the idx is removed in favor of an address (addr),
which is a relative address ranging over the UMEM. To convert an
idx-based address to the new addr is simply: addr = idx * frame_size +
offset.
We also stop referring to the UMEM "frame" as a frame. Instead it is
simply called a chunk.
To transfer ownership of a chunk to the kernel, the addr of the chunk
is passed in the fill-ring. Note, that the kernel will mask addr to
make it chunk aligned, so there is no need for userspace to do
that. E.g., for a chunk size of 2k, passing an addr of 2048, 2050 or
3000 to the fill-ring will refer to the same chunk.
On the completion-ring, the addr will match that of the Tx descriptor,
passed to the kernel.
Changing the descriptor format to use chunks/addr will allow for
future changes to move to a type-writer based model, where multiple
frames can reside in one chunk. In this model passing one single chunk
into the fill-ring, would potentially result in multiple Rx
descriptors.
This commit changes the uapi of AF_XDP sockets, and updates the
documentation.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Previously, rx_dropped could be updated incorrectly, e.g. if the XDP
program redirected the frame to a socket bound to a different queue
than where the XDP program was executing.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Previously the fill queue descriptor was not copied to kernel space
prior validating it, making it possible for userland to change the
descriptor post-kernel-validation.
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
No reason to use hard-coded protection information checks in ib_isert
driver. Use check masks from RDMA core driver.
Also, while we here, reduce the number of instructions made for setting
the check mask (no need to do bitwise or with 0 since we zero the mask
in the beginning of the function).
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Max Gurtovoy <maxg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
No reason to re-define protection information check in ib_iser driver.
Use check masks from RDMA core driver.
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Max Gurtovoy <maxg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
