- 11 Oct, 2023 3 commits
-
-
Gustavo A. R. Silva authored
Prepare for the coming implementation by GCC and Clang of the __counted_by attribute. Flexible array members annotated with __counted_by can have their accesses bounds-checked at run-time via CONFIG_UBSAN_BOUNDS (for array indexing) and CONFIG_FORTIFY_SOURCE (for strcpy/memcpy-family functions). Also, relocate `event->datalen = datalen;` to before calling `memcpy(event->data, data, datalen);`, so that the __counted_by annotation has effect, and flex-array member `data` can be properly bounds-checked at run-time. While there, use struct_size() helper, instead of the open-coded version, to calculate the size for the allocation of the whole flexible structure, including of course, the flexible-array member. This code was found with the help of Coccinelle, and audited and fixed manually. Signed-off-by:
Gustavo A. R. Silva <gustavoars@kernel.org> Reviewed-by:
Kees Cook <keescook@chromium.org> Acked-by:
Arend van Spriel <arend.vanspriel@broadcom.com> Signed-off-by:
Kalle Valo <kvalo@kernel.org> Link: https://lore.kernel.org/r/ZSRzrIe0345eymk2@work
-
Gustavo A. R. Silva authored
Prepare for the coming implementation by GCC and Clang of the __counted_by attribute. Flexible array members annotated with __counted_by can have their accesses bounds-checked at run-time via CONFIG_UBSAN_BOUNDS (for array indexing) and CONFIG_FORTIFY_SOURCE (for strcpy/memcpy-family functions). While there, use struct_size() helper, instead of the open-coded version, to calculate the size for the allocation of the whole flexible structure, including of course, the flexible-array member. This code was found with the help of Coccinelle, and audited and fixed manually. Signed-off-by:
-
Jinjie Ruan authored
Both usb_kill_urb() and usb_free_urb() do the NULL check itself, so there is no need to duplicate it prior to calling. Fixes: a82dfd33 ("wifi: rtw88: Add common USB chip support") Signed-off-by:
Jinjie Ruan <ruanjinjie@huawei.com> Acked-by:
Sascha Hauer <s.hauer@pengutronix.de> Acked-by:
Ping-Ke Shih <pkshih@realtek.com> Signed-off-by:
-
- 09 Oct, 2023 17 commits
-
-
Dmitry Antipov authored
Since 'bt_inq_page_start_time' of 'struct bt_coexist_8723' is in jiffies, prefer 'unsigned long' over 'u32' to avoid possible truncation in 'rtl8723e_dm_bt_inq_page_monitor()' and adjust related code. Found with clang's -Wshorten-64-to-32, compile tested only. Signed-off-by:
Dmitry Antipov <dmantipov@yandex.ru> Acked-by:
-
Zong-Zhe Yang authored
Since new criterion released by SRRC (State Radio Regulatory Commission, China) is stricter, we have adjusted TX power limit tables for it. But, due to RTL8821C HW characteristic, we still need to use specific parameter in CCK TX filter when set channel to avoid violations in some corner cases. Signed-off-by:
Zong-Zhe Yang <kevin_yang@realtek.com> Signed-off-by:
-
Zong-Zhe Yang authored
Sync Realtek Regulatory R42 and Realtek Channel Plan R64. Start to configure with Realtek regd CHILE, CN, UK, QATAR, UKRAINE. Signed-off-by:
-
Zong-Zhe Yang authored
Update TX power limit to parameter package V70 * tweak values of CN for its new regulation * configure values for QATAR, UK Signed-off-by:
-
Zong-Zhe Yang authored
Update TX power limit to parameter package V67 * configure values for MEXICO, CN, QATAR, UK Signed-off-by:
-
Zong-Zhe Yang authored
In newer Realtek parameter package, Realtek regd can configure QATAR and UK individually. So, driver extends the regd enum. Besides, driver configure alternative of them which will be referenced when parameter package of a chip doesn't consider QATAR and UK individually. Signed-off-by:
-
Felipe Negrelli Wolter authored
When frames are sent over the air, the device always applies the data rates in descending order. The driver assumed Minstrel also provided rate in descending order. However, in some cases, Minstrel can a choose a fallback rate greater than the primary rate. In this case, the two rates was inverted, the device try highest rate first and we get many retries. Since the device always applies rates in descending order, the workaround is to drop the rate when it higher than its predecessor in the rate list. Thus [ 4, 5, 3 ] becomes [ 4, 3 ]. This patch has been tested in isolated room with a series of attenuators. Here are the Minstrel statistics with 80dBm of attenuation: Without the fix: best ____________rate__________ ____statistics___ _____last____ ______sum-of________ mode guard # rate [name idx airtime max_tp] [avg(tp) avg(prob)] [retry|suc|att] [#success | #attempts] HT20 LGI 1 S MCS0 0 1477 5.6 5.2 82.7 3 0 0 3 4 HT20 LGI 1 MCS1 1 738 10.6 0.0 0.0 0 0 0 0 1 HT20 LGI 1 D MCS2 2 492 14.9 13.5 81.5 5 0 0 5 9 HT20 LGI 1 C MCS3 3 369 18.8 17.6 84.3 5 0 0 76 96 HT20 LGI 1 A P MCS4 4 246 25.4 22.4 79.5 5 0 0 11268 14026 HT20 LGI 1 B S MCS5 5 185 30.7 19.7 57.7 5 8 9 3918 9793 HT20 LGI 1 MCS6 6 164 33.0 0.0 0.0 5 0 0 6 102 HT20 LGI 1 MCS7 7 148 35.1 0.0 0.0 0 0 0 0 44 With the fix: best ____________rate__________ ____statistics___ _____last____ ______sum-of________ mode guard # rate [name idx airtime max_tp] [avg(tp) avg(prob)] [retry|suc|att] [#success | #attempts] HT20 LGI 1 S MCS0 0 1477 5.6 1.8 28.6 1 0 0 1 5 HT20 LGI 1 DP MCS1 1 738 10.6 9.7 82.6 4 0 0 14 34 HT20 LGI 1 MCS2 2 492 14.9 9.2 55.4 5 0 0 52 77 HT20 LGI 1 B S MCS3 3 369 18.8 15.6 74.9 5 1 1 417 554 HT20 LGI 1 A MCS4 4 246 25.4 16.7 59.2 5 1 1 13812 17951 HT20 LGI 1 C S MCS5 5 185 30.7 14.0 41.0 5 1 5 57 640 HT20 LGI 1 MCS6 6 164 33.0 0.0 0.0 0 0 1 0 48 HT20 LGI 1 S MCS7 7 148 35.1 0.0 0.0 0 0 0 0 36 We can notice the device try now to send with lower rates (and high success rates). At the end, we measured 20-25% better throughput with this patch. Fixes: 9bca45f3 ("staging: wfx: allow to send 802.11 frames") Tested-by:Olivier Souloumiac <olivier.souloumiac@silabs.com> Tested-by:
Alexandr Suslenko <suslenko.o@ajax.systems> Reported-by:
Jérôme Pouiller <jerome.pouiller@silabs.com> Signed-off-by:
Felipe Negrelli Wolter <felipe.negrelliwolter@silabs.com> Signed-off-by:
-
Jérôme Pouiller authored
With some conditions, the device is able to send/receive frames during scan operation. So, it is possible to use it implement the "remain on channel" feature. We just ask for a passive scan (without sending any probe request) on one channel. This architecture allows to leverage some interesting features: - if the device is AP, the device switches channel just after the next beacon and the beacons are stopped during the off-channel interval. - if the device is connected, it advertises it is asleep before to switch channel (so the AP should stop to try to send data) Signed-off-by: -
Jérôme Pouiller authored
Until now, all the traffic was blocked during scan operation. However, scan operation is going to be used to implement Remain On Channel (ROC). In this case, special frames (marked with IEEE80211_TX_CTL_TX_OFFCHAN) must be sent during the operation. These frames need to be sent on the virtual interface #2. Until now, this interface was only used by the device for internal purpose. But since API 3.9, it can be used to send data during scan operation (we hijack the scan process to implement ROC). Thus, we need to change a bit the way we match the frames with the interface. Fortunately, the frames received during the scan are marked with the correct interface number. So there is no change to do on this part. Signed-off-by:
-
Jérôme Pouiller authored
Currently, one scan_lock is associated to each vif. However, concurrent scan on vifs is explicitly prohibited by the device. Currently, scan_lock is associated with a vif but it is always locked with conf_mutex (there is a case where conf_mutex is not associated to scan_lock but scan_lock is tested on all interfaces). So concurrent scan on vifs cannot happen. So, this patch relocate scan_lock to the device and simplify the code. Signed-off-by:
-
Jérôme Pouiller authored
The device ignore the rx filters during the scan operation. wfx_configure_filter() acquires scan_lock to reflect this restriction. However, it is not really necessary since mac80211 don't try to configure Rx filters during scan. However, the things are changing. The scan operation is going to be used to implement remain-on-channel. In this case, wfx_configure_filter() can be called during the scan. Currently, this scenario generate a delay that end with a timeout in the upper layers. For the final user, some scenario of the EasyConnect specification end with a failure. So, avoid acquiring the scan_lock and just return. Signed-off-by:
-
Jérôme Pouiller authored
Like hof_scan(), hif_scan_uniq() invoke HIF_SCAN. However, it only allows to probe one channel and disable probe requests. It works very well to implement Remain-On-Channel. Signed-off-by:
-
Jérôme Pouiller authored
There is no real reasons to keep these function in the header file. Signed-off-by:
-
Jérôme Pouiller authored
wfx_rate_mask_to_hw() is only used in hif_tx.c. So relocate it into hif_tx.c and mark it static. Signed-off-by:
-
Jérôme Pouiller authored
The WF200 allow to start two network interfaces (one AP, one station) on two different channels. Since magic does not exist, it only works if the station interface enables power save. Thus, the driver detects this case and enforce power save as necessary. This patch fixes the case where the AP interface is stopped and it is no more necessary to enforce power saving on the station interface. Signed-off-by:
-
git://git.kernel.org/pub/scm/linux/kernel/git/kvalo/ath.gitKalle Valo authored
ath.git patches for v6.7. Major changes: ath12k * read board data variant name from SMBIOS
-
https://github.com/nbd168/wirelessKalle Valo authored
mt76 patches for 6.7 * mt7603/mt7628 stability improvements * fixes * new driver for mt7925
-
- 06 Oct, 2023 20 commits
-
-
Eric Dumazet authored
Exit early if the list is empty. Some applications using TCP zerocopy are calling recvmsg( ... MSG_ERRQUEUE) and hit this case quite often, probably because busy polling only deals with sk_receive_queue. Signed-off-by:
Eric Dumazet <edumazet@google.com> Link: https://lore.kernel.org/r/20231005114504.642589-1-edumazet@google.comSigned-off-by:
Jakub Kicinski <kuba@kernel.org>
-
Jakub Kicinski authored
Merge tag 'wireless-next-2023-10-06' of git://git.kernel.org/pub/scm/linux/kernel/git/wireless/wireless-next Kalle Valo says: ==================== wireless-next patches for v6.7 The first pull request for v6.7, with both stack and driver changes. We have a big change how locking is handled in cfg80211 and mac80211 which removes several locks and hopefully simplifies the locking overall. In drivers rtw89 got MCC support and smaller features to other active drivers but nothing out of ordinary. Major changes: cfg80211 - remove wdev mutex, use the wiphy mutex instead - annotate iftype_data pointer with sparse - first kunit tests, for element defrag - remove unused scan_width support mac80211 - major locking rework, remove several locks like sta_mtx, key_mtx etc. and use the wiphy mutex instead - remove unused shifted rate support - support antenna control in frame injection (requires driver support) - convert RX_DROP_UNUSABLE to more detailed reason codes rtw89 - TDMA-based multi-channel concurrency (MCC) support iwlwifi - support set_antenna() operation - support frame injection antenna control ath12k - WCN7850: enable 320 MHz channels in 6 GHz band - WCN7850: hardware rfkill support - WCN7850: enable IEEE80211_HW_SINGLE_SCAN_ON_ALL_BANDS to make scan faster ath11k - add chip id board name while searching board-2.bin * tag 'wireless-next-2023-10-06' of git://git.kernel.org/pub/scm/linux/kernel/git/wireless/wireless-next: (272 commits) wifi: rtlwifi: remove unreachable code in rtl92d_dm_check_edca_turbo() wifi: rtw89: debug: txpwr table supports Wi-Fi 7 chips wifi: rtw89: debug: show txpwr table according to chip gen wifi: rtw89: phy: set TX power RU limit according to chip gen wifi: rtw89: phy: set TX power limit according to chip gen wifi: rtw89: phy: set TX power offset according to chip gen wifi: rtw89: phy: set TX power by rate according to chip gen wifi: rtw89: mac: get TX power control register according to chip gen wifi: rtlwifi: use unsigned long for rtl_bssid_entry timestamp wifi: rtlwifi: fix EDCA limit set by BT coexistence wifi: rt2x00: fix MT7620 low RSSI issue wifi: rtw89: refine bandwidth 160MHz uplink OFDMA performance wifi: rtw89: refine uplink trigger based control mechanism wifi: rtw89: 8851b: update TX power tables to R34 wifi: rtw89: 8852b: update TX power tables to R35 wifi: rtw89: 8852c: update TX power tables to R67 wifi: rtw89: regd: configure Thailand in regulation type wifi: mac80211: add back SPDX identifier wifi: mac80211: fix ieee80211_drop_unencrypted_mgmt return type/value wifi: rtlwifi: cleanup few rtlxxxx_set_hw_reg() routines ... ==================== Link: https://lore.kernel.org/r/87jzrz6bvw.fsf@kernel.orgSigned-off-by:
-
Giulio Benetti authored
This patch adds the BCM5221 PHY support by reusing brcm_fet_*() callbacks and adding quirks for BCM5221 when needed. Cc: Jim Reinhart <jimr@tekvox.com> Cc: James Autry <jautry@tekvox.com> Cc: Matthew Maron <matthewm@tekvox.com> Signed-off-by:
Giulio Benetti <giulio.benetti+tekvox@benettiengineering.com> Reviewed-by:
Florian Fainelli <florian.fainelli@broadcom.com> Link: https://lore.kernel.org/r/20231005182915.153815-1-giulio.benetti@benettiengineering.comSigned-off-by:
-
git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queueJakub Kicinski authored
Tony Nguyen says: ==================== i40e: House-keeping and clean-up Ivan Vecera says: The series makes some house-keeping tasks on i40e driver: Patch 1: Removes unnecessary back pointer from i40e_hw Patch 2: Moves I40E_MASK macro to i40e_register.h where is used Patch 3: Refactors I40E_MDIO_CLAUSE* to use the common macro Patch 4: Add header dependencies to <linux/avf/virtchnl.h> Patch 5: Simplifies memory alloction functions Patch 6: Moves mem alloc structures to i40e_alloc.h Patch 7: Splits i40e_osdep.h to i40e_debug.h and i40e_io.h Patch 8: Removes circular header deps, fixes and cleans headers Patch 9: Moves DDP specific macros and structs to i40e_ddp.c * '40GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queue: i40e: Move DDP specific macros and structures to i40e_ddp.c i40e: Remove circular header dependencies and fix headers i40e: Split i40e_osdep.h i40e: Move memory allocation structures to i40e_alloc.h i40e: Simplify memory allocation functions virtchnl: Add header dependencies i40e: Refactor I40E_MDIO_CLAUSE* macros i40e: Move I40E_MASK macro to i40e_register.h i40e: Remove back pointer from i40e_hw structure ==================== Link: https://lore.kernel.org/r/20231005162850.3218594-1-anthony.l.nguyen@intel.comSigned-off-by:
-
Justin Stitt authored
`strncpy` is deprecated for use on NUL-terminated destination strings [1] and as such we should prefer more robust and less ambiguous string interfaces. We expect netdev->name to be NUL-terminated based on its use with format strings and dev_info(): | dev_info(&adapter->pdev->dev, | "%s link is up %d Mbps %s\n", | netdev->name, adapter->link_speed, | adapter->link_duplex == FULL_DUPLEX ? | "full duplex" : "half duplex"); Furthermore, NUL-padding is not required as netdev is already zero-initialized through alloc_etherdev(). Considering the above, a suitable replacement is `strscpy` [2] due to the fact that it guarantees NUL-termination on the destination buffer without unnecessarily NUL-padding. Link: https://www.kernel.org/doc/html/latest/process/deprecated.html#strncpy-on-nul-terminated-strings [1] Link: https://manpages.debian.org/testing/linux-manual-4.8/strscpy.9.en.html [2] Link: https://github.com/KSPP/linux/issues/90Signed-off-by:
Justin Stitt <justinstitt@google.com> Reviewed-by:
-
Justin Stitt authored
`strncpy` is deprecated for use on NUL-terminated destination strings [1] and as such we should prefer more robust and less ambiguous string interfaces. A suitable replacement is `strscpy` [2] due to the fact that it guarantees NUL-termination on the destination buffer without unnecessarily NUL-padding. It should be noted that there doesn't currently exist a bug here as DRV_NAME is a small string literal which means no overread bugs are present. Also to note, other ethernet drivers are using strscpy in a similar pattern: | dec/tulip/tulip_core.c | 861: strscpy(info->driver, DRV_NAME, sizeof(info->driver)); | | 8390/ax88796.c | 582: strscpy(info->driver, DRV_NAME, sizeof(info->driver)); | | dec/tulip/dmfe.c | 1077: strscpy(info->driver, DRV_NAME, sizeof(info->driver)); | | 8390/etherh.c | 558: strscpy(info->driver, DRV_NAME, sizeof(info->driver)); Link: https://www.kernel.org/doc/html/latest/process/deprecated.html#strncpy-on-nul-terminated-strings [1] Link: https://manpages.debian.org/testing/linux-manual-4.8/strscpy.9.en.html [2] Link: https://github.com/KSPP/linux/issues/90Signed-off-by:
Lukasz Stelmach <l.stelmach@samsung.com> Link: https://lore.kernel.org/r/20231005-strncpy-drivers-net-ethernet-asix-ax88796c_ioctl-c-v1-1-6fafdc38b170@google.comSigned-off-by:
-
Linus Walleij authored
As we don't specify the MTU in the driver, the framework will fall back to 1500 bytes and this doesn't work very well when we try to attach a DSA switch: eth1: mtu greater than device maximum ixp4xx_eth c800a000.ethernet eth1: error -22 setting MTU to 1504 to include DSA overhead After locating an out-of-tree patch in OpenWrt I found suitable code to set the MTU on the interface and ported it and updated it. Now the MTU gets set properly. Reviewed-by:
Jacob Keller <jacob.e.keller@intel.com> Signed-off-by:
Linus Walleij <linus.walleij@linaro.org> Link: https://lore.kernel.org/r/20231005-ixp4xx-eth-mtu-v4-1-08c66ed0bc69@linaro.orgSigned-off-by:
-
Harini Katakam authored
Since there's no alternate driver, change this entry from obsolete to orphan. Signed-off-by:
Harini Katakam <harini.katakam@amd.com> Link: https://lore.kernel.org/r/20231005131039.25881-1-harini.katakam@amd.comSigned-off-by:
-
Jakub Kicinski authored
Merge tag 'linux-can-next-for-6.7-20231005' of git://git.kernel.org/pub/scm/linux/kernel/git/mkl/linux-can-next Marc Kleine-Budde says: ==================== pull-request: can-next 2023-10-05 The first patch is by Miquel Raynal and fixes a comment in the sja1000 driver. Vincent Mailhol contributes 2 patches that fix W=1 compiler warnings in the etas_es58x driver. Jiapeng Chong's patch removes an unneeded NULL pointer check before dev_put() in the CAN raw protocol. A patch by Justin Stittreplaces a strncpy() by strscpy() in the peak_pci sja1000 driver. The next 5 patches are by me and fix the can_restart() handler and replace BUG_ON()s in the CAN dev helpers with proper error handling. The last 27 patches are also by me and target the at91_can driver. First a new helper function is introduced, the at91_can driver is cleaned up and updated to use the rx-offload helper. * tag 'linux-can-next-for-6.7-20231005' of git://git.kernel.org/pub/scm/linux/kernel/git/mkl/linux-can-next: (37 commits) can: at91_can: switch to rx-offload implementation can: at91_can: at91_alloc_can_err_skb() introduce new function can: at91_can: at91_irq_err_line(): send error counters with state change can: at91_can: at91_irq_err_line(): make use of can_change_state() and can_bus_off() can: at91_can: at91_irq_err_line(): take reg_sr into account for bus off can: at91_can: at91_irq_err_line(): make use of can_state_get_by_berr_counter() can: at91_can: at91_irq_err(): rename to at91_irq_err_line() can: at91_can: at91_irq_err_frame(): move next to at91_irq_err() can: at91_can: at91_irq_err_frame(): call directly from IRQ handler can: at91_can: at91_poll_err(): increase stats even if no quota left or OOM can: at91_can: at91_poll_err(): fold in at91_poll_err_frame() can: at91_can: add CAN transceiver support can: at91_can: at91_open(): forward request_irq()'s return value in case or an error can: at91_can: at91_chip_start(): don't disable IRQs twice can: at91_can: at91_set_bittiming(): demote register output to debug level can: at91_can: rename struct at91_priv::{tx_next,tx_echo} to {tx_head,tx_tail} can: at91_can: at91_setup_mailboxes(): update comments can: at91_can: add more register definitions can: at91_can: MCR Register: convert to FIELD_PREP() can: at91_can: MSR Register: convert to FIELD_PREP() ... ==================== Link: https://lore.kernel.org/r/20231005195812.549776-1-mkl@pengutronix.deSigned-off-by:
-
Johannes Berg authored
Resolve several conflicts, mostly between changes/fixes in wireless and the locking rework in wireless-next. One of the conflicts actually shows a bug in wireless that we'll want to fix separately. Signed-off-by:
Johannes Berg <johannes.berg@intel.com> Signed-off-by:
-
Sascha Hauer authored
This implements the led_hw_* hooks to support hardware blinking LEDs on the DP83867 phy. The driver supports all LED modes that have a corresponding TRIGGER_NETDEV_* define. Error and collision do not have a TRIGGER_NETDEV_* define, so these modes are currently not supported. Reviewed-by:
Andrew Lunn <andrew@lunn.ch> Tested-by: Alexander Stein <alexander.stein@ew.tq-group.com> #TQMa8MxML/MBa8Mx Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Kees Cook authored
Prepare for the coming implementation by GCC and Clang of the __counted_by attribute. Flexible array members annotated with __counted_by can have their accesses bounds-checked at run-time via CONFIG_UBSAN_BOUNDS (for array indexing) and CONFIG_FORTIFY_SOURCE (for strcpy/memcpy-family functions). As found with Coccinelle[1], add __counted_by for struct flow_action_entry. Cc: "David S. Miller" <davem@davemloft.net> Cc: Eric Dumazet <edumazet@google.com> Cc: Jakub Kicinski <kuba@kernel.org> Cc: Paolo Abeni <pabeni@redhat.com> Cc: netdev@vger.kernel.org Link: https://github.com/kees/kernel-tools/blob/trunk/coccinelle/examples/counted_by.cocci [1] Signed-off-by:
-
Kees Cook authored
Prepare for the coming implementation by GCC and Clang of the __counted_by attribute. Flexible array members annotated with __counted_by can have their accesses bounds-checked at run-time via CONFIG_UBSAN_BOUNDS (for array indexing) and CONFIG_FORTIFY_SOURCE (for strcpy/memcpy-family functions). As found with Coccinelle[1], add __counted_by for struct packet_fanout. Cc: "David S. Miller" <davem@davemloft.net> Cc: Eric Dumazet <edumazet@google.com> Cc: Jakub Kicinski <kuba@kernel.org> Cc: Paolo Abeni <pabeni@redhat.com> Cc: Willem de Bruijn <willemb@google.com> Cc: Anqi Shen <amy.saq@antgroup.com> Cc: netdev@vger.kernel.org Link: https://github.com/kees/kernel-tools/blob/trunk/coccinelle/examples/counted_by.cocci [1] Signed-off-by:
-
Jakub Kicinski authored
Chuck points out that we should use the uapi-header property when generating the guard. Otherwise we may generate the same guard as another file in the tree. Tested-by:
Chuck Lever <chuck.lever@oracle.com> Signed-off-by:
Przemek Kitszel <przemyslaw.kitszel@intel.com> Signed-off-by:
-
David S. Miller authored
Petr Machata says: ==================== mlxsw: Control the order of blocks in ACL region Amit Cohen writes: For 12 key blocks in the A-TCAM, rules are split into two records, which constitute two lookups. The two records are linked using a "large entry key ID". Due to a Spectrum-4 hardware issue, KVD entries that correspond to key blocks 0 to 5 of 12 key blocks will be placed in the same KVD pipe if they only differ in their "large entry key ID", as it is ignored. This results in a reduced scale, we can insert less than 20k filters and get an error: $ tc -b flower.batch RTNETLINK answers: Input/output error We have an error talking to the kernel To reduce the probability of this issue, we can place key blocks with high entropy in blocks 0 to 5. The idea is to place blocks that are often changed in blocks 0 to 5, for example, key blocks that match on IPv4 addresses or the LSBs of IPv6 addresses. Such placement will reduce the probability of these blocks to be same. Mark several blocks with 'high_entropy' flag and place them in blocks 0 to 5. Note that the list of the blocks is just a suggestion, I will verify it with architects. Currently, there is a one loop that chooses which blocks should be used for a given list of elements and fills the blocks - when a block is chosen, it fills it in the region. To be able to control the order of the blocks, separate between searching blocks and filling them. Several pre-changes are required. Patch set overview: Patch #1 marks several blocks with 'high_entropy' flag. Patches #2-#4 prepare the code for filling blocks at the end of the search. Patch #5 changes the loop to just choose the blocks and fill the blocks at the end. ==================== Signed-off-by: -
Amit Cohen authored
The previous patches prepared the code to allow separating between choosing blocks and filling blocks. Do not add blocks as part of the loop that chooses them. When all the required blocks are set in the bitmap 'chosen_blocks_bm', start filling blocks. Iterate over the bitmap twice - first add only blocks that are marked with 'high_entropy' flag. Then, fill the rest of the blocks. The idea is to place key blocks with high entropy in blocks 0 to 5. See more details in previous patches. Signed-off-by:
Amit Cohen <amcohen@nvidia.com> Reviewed-by:
Ido Schimmel <idosch@nvidia.com> Signed-off-by:
Petr Machata <petrm@nvidia.com> Signed-off-by:
-
Amit Cohen authored
Currently, mlxsw_afk_picker() chooses which blocks will be used for a given list of elements, and fills the blocks during the searching - when a key block is found with most hits, it adds it and removes the elements from the count of hits. This should be changed as we want to be able to choose which blocks will be placed in blocks 0 to 5. To separate between choosing blocks and filling blocks, several pre-changes are required. Currently, the indication of whether all elements were found in the chosen blocks is by the structure 'key_info->elusage'. This structure is updated when block is filled as part of mlxsw_afk_picker_key_info_add(). A following patch will call this function only after choosing all the blocks. Add a bitmap called 'elusage_chosen' to store which elements were chosen in the chosen blocks. Change the condition in the loop to check elements that were chosen, not elements that were already filled in the blocks. Signed-off-by:
-
Amit Cohen authored
Currently, mlxsw_afk_picker() chooses which blocks will be used for a given list of elements, and fills the blocks during the searching - when a key block is found with most hits, it adds it and removes the elements from the count of hits. This should be changed as we want to be able to choose which blocks will be placed in blocks 0 to 5. To separate between choosing blocks and filling blocks, several pre-changes are required. During the search, the structure 'mlxsw_afk_picker' is used per block, it contains how many elements from the required list appear in the block. When a block is chosen and filled, this bitmap of elements is cleaned. To be able to fill the blocks at the end, add a bitmap called 'chosen_element' as part of picker. When a block is chosen, copy the 'element' bitmap to it. Use the new bitmap as part of mlxsw_afk_picker_key_info_add(). So later, when filling the block will be done at the end of the searching, we will use the copied bitmap that contains the elements that should be used in the block. Signed-off-by:
-
Amit Cohen authored
Currently, mlxsw_afk_picker() chooses which blocks will be used for a given list of elements, and fills the blocks during the searching - when a key block is found with most hits, it adds it and removes the elements from the count of hits. This should be changed as we want to be able to choose which blocks will be placed in blocks 0 to 5. To separate between choosing blocks and filling blocks, several pre-changes are required. The indexes of the chosen blocks should be saved, so then the relevant blocks will be filled at the end of search. Allocate a bitmap for chosen blocks, when a block is found with most hits, set the relevant bit in the bitmap. This bitmap will be used in a following patch. Signed-off-by:
-
Amit Cohen authored
For 12 key blocks in the A-TCAM, rules are split into two records, which constitute two lookups. The two records are linked using a "large entry key ID". Due to a Spectrum-4 hardware issue, KVD entries that correspond to key blocks 0 to 5 of 12 key blocks A-TCAM entries will be placed in the same KVD pipe if they only differ in their "large entry key ID", as it is ignored. This results in a reduced scale. To reduce the probability of this issue, we can place key blocks with high entropy in blocks 0 to 5. The idea is to place blocks that are changed often in blocks 0 to 5, for example, key blocks that match on IPv4 addresses or the LSBs of IPv6 addresses. Such placement will reduce the probability of these blocks to be same. Mark several blocks with 'high_entropy' flag, so later we will take into account this flag and place them in blocks 0 to 5. Signed-off-by:
-
