Generic Tx stats fixes:
- do not try to parse "aggregation status" for single frames; send them
to iwn_tx_done() instead;
- try to attach mbuf / node reference pair to reported BA events;
allows to fix reported status for ieee80211_tx_complete() and ifnet counters
(previously all A-MPDU frames were counted as failed - see PR 210211);
requires few more firmware bug workarounds;
- preserve short / long retry counters for wlan_amrr(4)
(disabled for now - causes significant performance degradation).
- Add new IWN_DEBUG_AMPDU debug category.
- Add one more check into iwn_tx_data() to prevent aggregation ring
overflow.
- Workaround 'seqno % 256' != 'current Tx slot' case (until D9195 is not
in the tree).
- Improve watchdog timer updates (previously watchdog check was omitted
when at least one frame was transmitted).
- Stop Tx when memory leak in currently used ring was detected (unlikely
to happen).
- Few other minor fixes.
Was previously tested with:
- Intel 6205, STA mode (Tx aggregation behaves much better now).
- Intel 4965AGN, STA mode (still unstable).
PR: 192641, 210211
Reviewed by: adrian, dhw
MFC after: 1 month
Differential Revision: https://reviews.freebsd.org/D10728
Rate tables have this bit set to indicate minimal set of basic rates;
however, it overlappes with MCS bit, so rate2ridx() will treat them as
an 11n rate.
Due to the current rates setup the issue can be reproduced only
in 5GHz band with 11n / protection enabled.
Tested with RTL8821AU, HOSTAP mode.
MFC after: 5 days
In FreeBSD, this is normal situation that the Tx ring is being full. In
hat case, the packet is put back into drbr and the next attempt to send
it is taken after the cleanup.
Too much logs like this can cause system instability and even cause the
device reset (because keep alive or cleanup could be missed).
To fix that, the log level of this message is changed to debug.
Upon this change upgrade the driver version to v0.8.2.
Submitted by: Michal Krawczyk <mk@semihalf.com>
Obtained from: Semihalf
Sponsored by: Amazon, Inc.
upstream it seems).
The tlv variable was changed to a pointer but the advancement of the data pointer
was left as sizeof(tlv). While the sizeof the (now) pointer equals the
sizeof 2 x uint32_t (size of the struct) on 64bit platforms, on 32bit platforms
the size of the advancement of the data pointer was wrong leading to
firmware load issues.
Correctly advance the data pointer by the size of the structure and not by
the size of a pointer.
PR: 219683
Submitted by: waddlesplash gamil.com (Haiku) on irc
MFC after: 1 week
ccr reuses the control queue and first rx queue from the first port on
each adapter. The driver cannot send requests until those queues are
initialized. Refuse to create sessions for now if the queues aren't
ready. This is a workaround until cxgbe allocates one or more
dedicated queues for ccr.
PR: 233851
MFC after: 1 week
Sponsored by: Chelsio Communications
Differential Revision: https://reviews.freebsd.org/D18478
o In vm_pager_bufferinit() create pbuf_zone and start accounting on how many
pbufs are we going to have set.
In various subsystems that are going to utilize pbufs create private zones
via call to pbuf_zsecond_create(). The latter calls uma_zsecond_create(),
and sets a limit on created zone. After startup preallocate pbufs according
to requirements of all pbuf zones.
Subsystems that used to have a private limit with old allocator now have
private pbuf zones: md(4), fusefs, NFS client, smbfs, VFS cluster, FFS,
swap, vnode pager.
The following subsystems use shared pbuf zone: cam(4), nvme(4), physio(9),
aio(4). They should have their private limits, but changing that is out of
scope of this commit.
o Fetch tunable value of kern.nswbuf from init_param2() and while here move
NSWBUF_MIN to opt_param.h and eliminate opt_swap.h, that was holding only
this option.
Default values aren't touched by this commit, but they probably should be
reviewed wrt to modern hardware.
This change removes a tight bottleneck from sendfile(2) operation, that
uses pbufs in vnode pager. Other pagers also would benefit from faster
allocation.
Together with: gallatin
Tested by: pho
Do not lose error condition by always returning 0 from set_led.
None of the calls to set_led checks for return value at the moment so
none of API consumers in base is affected.
PR: 231567
Submitted by: Bertrand Petit <bsdpr@phoe.frmug.org>
MFC after: 1 week
Family 15h is a bit of an oddball. Early models used the same temperature
register and spec (mostly[1]) as earlier CPU families.
Model 60h-6Fh and 70-7Fh use something more like Family 17h's Service
Management Network, communicating with it in a similar fashion. To support
them, add support for their version of SMU indirection to amdsmn(4) and use
it in amdtemp(4) on these models.
While here, clarify some of the deviceid macros in amdtemp(4) that were
added with arbitrary, incorrect family numbers, and remove ones that were
not used. Additionally, clarify intent and condition of heterogenous
multi-socket system detection.
[1]: 15h adds the "adjust range by -49°C if a certain condition is met,"
which previous families did not have.
Reported by: D. C. <tjoard AT gmail.com>
PR: 234657
Tested by: D. C. <tjoard AT gmail.com>
Extend the vendor class USB audio quirk to cover devices without
the USB audio control descriptor.
PR: 234794
MFC after: 1 week
Sponsored by: Mellanox Technologies
Issue:
ocs_fc(4) driver panics. It's induced by setting the port_state
sysctl to offline, then online, then offline, then online, and so
forth and so on in rapid succession.
Reason:
While we set the port_state to online fc discovery will start and OS
is enumerating the target discs by calling ocs_action(), then set the
port state to "offline" which deletes domain/sport/nodes.
In ocs_action()->XPT_GET_TRAN_SETTINGS we are accessing the remote
node which can be invalid to get the wwpn, wwnn and port.
Fix:
Removed accessing of remote node and domain in some ocs_action() cases.
Populated the required values from ocs_fcport.
This removes the dependency of node and domain structures while
processing XPT_PATH_INQ and XPT_GET_TRAN_SETTINGS.
We will invalidate the target entries after the device lost
timeout(30 seconds).
Approved by: ken, mav
MFC after: 3 weeks
The code is similar to the one for RTL8188E* and probably
should be shared with RTL8188CE (needs to be tested).
Checked with RTL8188CUS, STA mode.
MFC after: 5 days
- Remove macros that covertly create epoch_tracker on thread stack. Such
macros a quite unsafe, e.g. will produce a buggy code if same macro is
used in embedded scopes. Explicitly declare epoch_tracker always.
- Unmask interface list IFNET_RLOCK_NOSLEEP(), interface address list
IF_ADDR_RLOCK() and interface AF specific data IF_AFDATA_RLOCK() read
locking macros to what they actually are - the net_epoch.
Keeping them as is is very misleading. They all are named FOO_RLOCK(),
while they no longer have lock semantics. Now they allow recursion and
what's more important they now no longer guarantee protection against
their companion WLOCK macros.
Note: INP_HASH_RLOCK() has same problems, but not touched by this commit.
This is non functional mechanical change. The only functionally changed
functions are ni6_addrs() and ni6_store_addrs(), where we no longer enter
epoch recursively.
Discussed with: jtl, gallatin
Dell-branded Intel P4600 NVMe drives benefit from NVMe 1.3's NOIOB
feature. Unfortunately just like Intel DC P4500s, they don't advertise
themselves as benefiting from this...
This changes adds P4600s to the existing list of old drives which
benefit from striping.
PR: 233969
Submitted by: David Fugate <dave.fugate@gmail.com>
Reviewed by: imp, mav
Approved by: imp (mentor)
MFC after: 1 week
Differential Revision: https://reviews.freebsd.org/D18772
setting the data prior to setting up the interrupt. Now we only set
the cookie afterwards, and that (a) cannot be helpd and (b) isn't used
in the ISR.
PR: 147127
Submitted by: hps@
On system with Celeron 1.5GHz CPU, sometimes when a PCMCIA to Compact Flash
adapter containing a Compact Flash card is inserted in the cardbus slot the
system hangs. This problem has not been observed in systems with a 2.8GHz
XEON CPU or faster.
Analysis of the cbb driver shows functional interrupts are routed to PCI
BEFORE the interrupt handler for functional interrupts has been registered.
Fix applied as described in the bug.
PR: 128040
Submitted by: Arthur Hartwig
Use BUS_DMA_NOWAIT for loads at initialization time.
Report actual numeric error code if any problem occurs at the
initialization.
Reported and tested by: pho
Reviewed by: mav
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
Differential revision: https://reviews.freebsd.org/D18741
tws_passthru() was doing a copyin of a user-specified request
without validating its length, so a malicious request could overrun
the buffer. By default, the tws(4) device file is only accessible
as root.
admbug: 825
Reported by: Anonymous of the Shellphish Grill Team
Reviewed by: delphij
MFC after: 1 week
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D18536
out dead USB HUB devices by implementing an error counter, so that the USB
enumeration thread does not spend all its time reading from non-responding
devices, blocking user-space access in the end.
Tested by: Matthias Apitz <guru@unixarea.de>
MFC after: 1 week
Sponsored by: Mellanox Technologies
done on the old keyboard and then do the corresponding number of grabs
on the new keyboard.
This fixes a race that can leave the system with a non-functioning
keyboard. It goes like this...
- The bios claims there is an AT keyboard, atkbd attaches.
- SI_SUB_INT_CONFIG_HOOKS runs.
- USB probes devices. Devices begin attaching, including disks.
- GELI prompts for a password for a just-attached disk, which results
in a cngrab() while atkbd is the keyboard.
- A USB keyboard attaches.
- vt_upgrade() runs and switches the keyboard to the new USB keyboard,
but because cngrab was never called for it, it's not activated and
keystrokes are ignored.
- Now there is no functional keyboard and no way to get one; even
plugging in a different USB keyboard doesn't help, because the console
is still grabbed, still waiting for a GELI pw.
Discussed with: ray@
front-end doesn't support SDMA or the latter implements a platform-
specific transfer method instead. While at it, factor out allocation
and freeing of SDMA resources to sdhci_dma_{alloc,free}() in order to
keep the code more readable when adding support for ADMA variants.
o Base the size of the SDMA bounce buffer on MAXPHYS up to the maximum
of 512 KiB instead of using a fixed 4-KiB-buffer. With the default
MAXPHYS of 128 KiB and depending on the controller and medium, this
reduces the number of SDHCI interrupts by a factor of ~16 to ~32 on
sequential reads while an increase of throughput of up to ~84 % was
seen.
Front-ends for broken controllers that only support an SDMA buffer
boundary of a specific size may set SDHCI_QUIRK_BROKEN_SDMA_BOUNDARY
and supply a size via struct sdhci_slot. According to Linux, only
Qualcomm MSM-type SDHCI controllers are affected by this, though.
Requested by: Shreyank Amartya (unconditional bump to 512 KiB)
o Introduce a SDHCI_DEPEND macro for specifying the dependency of the
front-end modules on the sdhci(4) one and bump the module version
of sdhci(4) to 2 via an also newly introduced SDHCI_VERSION in order
to ensure that all components are in sync WRT struct sdhci_slot.
o In sdhci(4):
- Make pointers const were applicable,
- replace a few device_printf(9) calls with slot_printf() for
consistency, and
- sync some local functions with their prototypes WRT static.
Previous code typically crashed in case of NVMe device unplug or even clean
detach while some I/Os are still in flight. To fix this the new code calls
disk_gone() and waits for confirmation of all references gone before calling
disk_destroy(), freeing other resources and allowing controller detach.
While there, fix disk lists locking and reimplement unit numbers assignment.
MFC after: 1 month
Sponsored by: iXsystems, Inc.
