- 11 Jun, 2021 40 commits
-
-
Alex Elder authored
The cost of validating the endpoint configuration data is not all that high, so just do it unconditionally, rather than doing so only when IPA_VALIDATAION is defined. Suggested-by:
Leon Romanovsky <leon@kernel.org> Signed-off-by:
Alex Elder <elder@linaro.org> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Lijun Pan authored
drivers/net/ethernet/ibm/ibmvnic.c: In function ‘handle_vpd_rsp’: drivers/net/ethernet/ibm/ibmvnic.c:4393:3: warning: ‘strncpy’ output truncated before terminating nul copying 3 bytes from a string of the same length [-Wstringop-truncation] 4393 | strncpy((char *)adapter->fw_version, "N/A", 3 * sizeof(char)); | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Signed-off-by:Lijun Pan <lijunp213@gmail.com> Signed-off-by:
-
David S. Miller authored
Vladimir Oltean says: ==================== Port the SJA1105 DSA driver to XPCS As requested when adding support for the NXP SJA1110, the SJA1105 driver could make use of the common XPCS driver, to eliminate some hardware specific code duplication. This series modifies the XPCS driver so that it can accommodate the XPCS instantiation from NXP switches, and the SJA1105 driver so it can expose what the XPCS driver expects. Tested on NXP SJA1105S and SJA1110A. Changes in v3: None. This is a resend of v2 which had "changes requested" even though there was no direct feedback. ==================== Signed-off-by: -
Vladimir Oltean authored
The MAC treats 2500base-x same as SGMII (yay for that) except that it must be set to a different speed. Extend all places that check for SGMII to also check for 2500base-x. Also add the missing 2500base-x compatibility matrix entry for SJA1110D. Signed-off-by:
Vladimir Oltean <vladimir.oltean@nxp.com> Signed-off-by:
-
Vladimir Oltean authored
For the xMII Mode Parameters Table to be properly configured for SGMII mode on SJA1110, we need to set the "special" bit, since SGMII is officially bitwise coded as 0b0011 in SJA1105 (decimal 3, equal to XMII_MODE_SGMII), and as 0b1011 in SJA1110 (decimal 11). Signed-off-by:
-
Vladimir Oltean authored
On the SJA1110, the PCS of each SERDES-capable port is accessed through a different memory window which is 0x100 bytes in size, denoted by "pcs_base". In each PCS register access window, the XPCS MMDs are accessed in an indirect way: in pages/banks of up to 0x100 addresses each. Changing the page/bank is done by writing to a special register at the end of the access window. The MDIO register map accessed indirectly through the indirect banked method described above is similar to what SJA1105 has: upper 5 bits are the MMD, lower 16 bits are the MDIO address within that MMD. Since the PHY ID reported by the XPCS inside SJA1110 is also all zeroes (like SJA1105), we need to trap those reads and return a fake PHY ID so that the xpcs driver can apply some specific fixups for our integration. Signed-off-by:
-
Vladimir Oltean authored
There is a desire to use the generic driver for the Synopsys XPCS located in drivers/net/pcs, and to achieve that, the sja1105 driver must expose an MDIO bus for the SGMII PCS, because the XPCS probes as an mdio_device. In preparation of the SJA1110 which in fact has a different access procedure for the SJA1105, we register this PCS MDIO bus once in the common code, but we implement function pointers for the read and write methods. In this patch there is a single implementation for them. There is exactly one MDIO bus for the PCS, this will contain all PCSes at MDIO addresses equal to the port number. We delete a bunch of hardware support code because the xpcs driver already does what we need. We need to hack up the MDIO reads for the PHY ID, since our XPCS instantiation returns zeroes and there are some specific fixups which need to be applied by the xpcs driver. Signed-off-by:
-
Vladimir Oltean authored
The sja1105 hardware has a quirk in that some changes require a switch reset, which loses all configuration. When the reset is initiated, everything needs to be reprogrammed, including the MACs and the PCS. This is currently done in sja1105_static_config_reload() - we manually call sja1105_adjust_port_config(), sja1105_sgmii_pcs_config() and sja1105_sgmii_pcs_force_speed() which are all internal functions. There is a desire for sja1105 to use the common xpcs driver, and that means that the equivalents of those functions, xpcs_do_config() and xpcs_link_up() respectively, will no longer be local functions. Forcing phylink to retrigger a resolve somehow, say by doing dev_close() followed by dev_open() is not really an option, because the CPU port might have a PCS as well, and there is no net device which we can close and reopen for that. Additionally, the dev_close/dev_open sequence might force a renegotiation of the copper-side link for SGMII ports connected to a PHY, and this is undesirable as well, because the switch reset is much quicker than a PHY autoneg, so we would have a lot more downtime. The only solution I see is for the sja1105 driver to keep doing what it's doing, and that means we need to export the equivalents from xpcs for sja1105_sgmii_pcs_config and sja1105_sgmii_pcs_force_speed, and call them directly in sja1105_static_config_reload(). This will be done during the conversion patch. Signed-off-by:
-
Vladimir Oltean authored
The NXP SJA1110 switch integrates its own, non-Synopsys PMA, but it manages it through the register space of the XPCS itself, in a small register window inside MDIO_MMD_VEND2 from address 0x8030 to 0x806e. This coincides with where the registers for the default Synopsys PMA are, but the register definitions are of course not the same. This situation is an odd hardware quirk, but the simplest way to manage it is to drive the SJA1110's PMA from within the XPCS driver. Signed-off-by:
-
Vladimir Oltean authored
The NXP SJA1105 DSA switch integrates a Synopsys SGMII XPCS on port 4. The generic code works fine, except there is an integration issue which needs to be dealt with: in this switch, the XPCS is integrated with a PMA that has the TX lane polarity inverted by default (PLUS is MINUS, MINUS is PLUS). To obtain normal non-inverted behavior, the TX lane polarity must be inverted in the PCS, via the DIGITAL_CONTROL_2 register. We introduce a pma_config() method in xpcs_compat which is called by the phylink_pcs_config() implementation. Also, the NXP SJA1105 returns all zeroes in the PHY ID registers 2 and 3. We need to hack up an ad-hoc PHY ID (OUI is zero, device ID is 1) in order for the XPCS driver to recognize it. This PHY ID is added to the public include/linux/pcs/pcs-xpcs.h for that reason (for the sja1105 driver to be able to use it in a later patch). Signed-off-by:
-
Vladimir Oltean authored
xpcs_get_id() searches multiple MMDs for a known PHY ID, starting with MDIO_MMD_PCS (3). However not all integrators might have implemented that MMD on their MDIO bus. For example, the NXP SJA1105 and SJA1110 switches only implement vendor-specific MMD 1 and 2. When there is nothing on an MDIO bus at a certain address, traditionally the bus returns 0xffff, which means that the bus remained in its default pull-up state for the duration of the MDIO transaction. The 0xffff value is widely used in drivers/net/phy/phy_device.c (see get_phy_c22_id for example) to denote a missing device. So it makes sense for the xpcs to ignore this value as well, and continue its search, eventually finding the proper PHY ID in the vendor-specific MMDs. Signed-off-by:
-
Vladimir Oltean authored
In fixed-link use cases, the XPCS can disable the clause 37 in-band autoneg process, disable the "Automatic Speed Mode Change after CL37 AN" setting, and force operation in a speed dictated by management. Add support for this operating mode. Signed-off-by:
-
Vladimir Oltean authored
Vendors which integrate the Designware XPCS might modify a few things here and there, and to support those, it's best to create separate C files in order to not clutter up the main pcs-xpcs.c. Because the vendor files might want to access the common xpcs registers too, let's move them in a header file which is local to this driver and can be included by vendor files as appropriate. Signed-off-by:
-
Vladimir Oltean authored
There is no reason to embed an if within an if, we can just logically AND the two conditions. Signed-off-by:
Wong Vee Khee <vee.khee.wong@linux.intel.com> Signed-off-by:
-
Vladimir Oltean authored
Reorder the variable declarations in descending line length order, according to the networking coding style. Signed-off-by:
-
Vladimir Oltean authored
The struct mdio_xpcs_args is reminiscent of when a similarly named struct mdio_xpcs_ops existed. Now that that is removed, we can shorten the name to dw_xpcs (dw for DesignWare). Signed-off-by:
-
David S. Miller authored
Alex Elder says: ==================== net: qualcomm: rmnet: MAPv4 download checksum cleanup, part 1 I'm posting a large series an two smaller parts; this is part 1. The RMNet driver handles MAP (or QMAP) protocol traffic. There are several versions of this protocol. Version 1 supports multiplexing, as well as aggregation of packets in a single buffer. Version 4 adds the ability to perform checksum offload. And version 5 implements checksum offload in a different way from version 4. This series involves only MAPv4 protocol checksum offload, and only in the download (RX) direction. It affects handling of checksums computed by hardware for UDP datagrams and TCP segments, carried over both IPv4 and IPv6. MAP packets arriving on an RMNet port implementing MAPv4 checksum offload are passed to rmnet_map_checksum_downlink_packet() for handling. The packet is then passed to rmnet_map_ipv4_dl_csum_trailer() or rmnet_map_ipv6_dl_csum_trailer(), depending contents of the MAP payload. These two functions interpret checksum metadata to determine whether the checksum in the received packet matches that calculated by the hardware. It is these two functions that are the subject of this series (parts 1 and 2). The bulk of these functions are transformed--in a lot of small steps--from an extremely difficult-to-follow block of checksum processing code into a fairly simple, heavily commented equivalent. ==================== Signed-off-by: -
Alex Elder authored
In the previous patch IPv4 download checksum offload code was updated to avoid unnecessary byte swapping, based on properties of the Internet checksum algorithm. This patch makes comparable changes to the IPv6 download checksum offload handling. Signed-off-by:
-
Alex Elder authored
Internet checksums are used for IPv4 header checksum, as well as TCP segment and UDP datagram checksums. Such a checksum represents the negated sum of adjacent pairs of bytes, using ones' complement arithmetic. One property of the Internet checkum is byte order independence [1]. Specifically, the sum of byte-swapped pairs is equal to the result of byte swapping the sum of those same pairs when not byte-swapped. So for example if a, b, c, d, y, and z are hexadecimal digits, and PLUS represents ones' complement addition: If: ab PLUS cd = yz Then: ba PLUS dc = zy For this reason, there is no need to swap the order of bytes in the checksum value held in a message header, nor the one in the QMAPv4 trailer, in order to operate on them. In other words, we can determine whether the hardware-computed checksum matches the one in the message header without any byte swaps. (This patch leaves in place all existing type casts.) [1] https://tools.ietf.org/html/rfc1071Signed-off-by: -
Alex Elder authored
In rmnet_map_ipv6_dl_csum_trailer() there is an especially involved line of code that determines the ones' complement sum of the IPv6 packet header (in host byte order). Simplify that by storing the result of computing just the header checksum in a local variable, then using that in the original assignment. Use the size of the IPv6 header structure as the number of bytes to checksum, rather than computing the offset to the transport header. And use ip_fast_csum() rather than ipa_compute_csum(), knowing that the size of an IPv6 header (40 bytes) is a multiple of 4 bytes greater than 16. Add some comments to match rmnet_map_ipv4_dl_csum_trailer(). Signed-off-by:
-
Alex Elder authored
In rmnet_map_ipv4_dl_csum_trailer(), an illegal checksum subtraction is done, subtracting hdr_csum (in host byte order) from csum_value (in network byte order). Despite being illegal, it generally works, because it turns out the value subtracted is (or should be) always 0, which has the same representation in either byte order. Doing illegal operations is not good form though, so fix this by verifying the IP header checksum early in that function. If its checksum is non-zero, the packet will be bad, so just return an error. This will cause the packet to passed to the IP layer where it can be dropped. Thereafter, there is no need subtract the IP header checksum from the checksum value in the trailer because we know it is zero. Add a comment explaining this. This type of packet error is different from other types, so add a new statistics counter to track this condition. Signed-off-by:
-
Alex Elder authored
The checksum fields of the TCP and UDP header structures already have type __sum16. We don't support any other protocol headers, so we can simplify rmnet_map_get_csum_field(), getting rid of the local variable entirely and just returning the appropriate address. Signed-off-by:
-
Alex Elder authored
The value passed as an argument to rmnet_map_ipv4_ul_csum_header() is always an IPv4 header. Rather than using a local variable, just have the type of the argument reflect the proper type. In rmnet_map_ipv6_ul_csum_header() things are defined a little differently, but make the same basic change there. Signed-off-by:
-
Alex Elder authored
If IPV6 is not enabled in the kernel configuration, the RMNet checksum code indicates a buffer containing an IPv6 packet is not supported. The same thing happens if a buffer contains something other than an IPv4 or IPv6 packet. We can rearrange things a bit in two functions so that some #ifdef calls can simply be eliminated. Signed-off-by:
-
Alex Elder authored
In rmnet_map_ipv4_dl_csum_trailer() use ip_is_fragment() to determine whether a socket buffer contains a packet fragment. Signed-off-by:
-
git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queueDavid S. Miller authored
Tony Nguyen says: ==================== Jake Keller says: ==================== 100GbE Intel Wired LAN Driver Updates 2021-06-11 Extend the ice driver to support basic PTP clock functionality for E810 devices. This includes some tangential work required to setup the sideband queue and driver shared parameters as well. This series only supports E810-based devices. This is because other devices based on the E822 MAC use a different and more complex PHY. The low level device functionality is kept within ice_ptp_hw.c and is designed to be extensible for supporting E822 devices in a future series. This series also only supports very basic functionality including the ptp_clock device and timestamping. Support for configuring periodic outputs and external input timestamps will be implemented in a future series. There are a couple of potential "what? why?" bits in this series I want to point out: 1) the PTP hardware functionality is shared between multiple functions. This means that the same clock registers are shared across multiple PFs. In order to avoid contention or clashing between PFs, firmware assigns "ownership" to one PF, while other PFs are merely "associated" with the timer. Because we share the hardware resource, only the clock owner will allocate and register a PTP clock device. Other PFs determine the appropriate PTP clock index to report by using a firmware interface to read a shared parameter that is set by the owning PF. 2) the ice driver uses its own kthread instead of using do_aux_work. This is because the periodic and asynchronous tasks are necessary for all PFs, but only one PF will allocate the clock. The series is broken up into functional pieces to allow easy review. ==================== Signed-off-by: -
David S. Miller authored
Arseny Krasnov says: ==================== virtio/vsock: introduce SOCK_SEQPACKET support This patchset implements support of SOCK_SEQPACKET for virtio transport. As SOCK_SEQPACKET guarantees to save record boundaries, so to do it, new bit for field 'flags' was added: SEQ_EOR. This bit is set to 1 in last RW packet of message. Now as packets of one socket are not reordered neither on vsock nor on vhost transport layers, such bit allows to restore original message on receiver's side. If user's buffer is smaller than message length, when all out of size data is dropped. Maximum length of datagram is limited by 'peer_buf_alloc' value. Implementation also supports 'MSG_TRUNC' flags. Tests also implemented. Thanks to stsp2@yandex.ru for encouragements and initial design recommendations. ==================== Signed-off-by: -
Arseny Krasnov authored
Add SEQPACKET socket type to vsock trace event. Signed-off-by:
Arseny Krasnov <arseny.krasnov@kaspersky.com> Signed-off-by:
-
Arseny Krasnov authored
Implement two tests of SOCK_SEQPACKET socket: first sends data by several 'write()'s and checks that number of 'read()' were same. Second test checks MSG_TRUNC flag. Cases for connect(), bind(), etc. are not tested, because it is same as for stream socket. Signed-off-by:
-
Arseny Krasnov authored
Add SEQPACKET ops for loopback transport and 'seqpacket_allow()' callback. Signed-off-by:
-
Arseny Krasnov authored
When received packet is copied to guests's rx queue, data buffers of rx queue could be smaller that data buffer of input packet, so data of input packet is copied to each rx buffer, thus each rx buffer will be a packet with dynamically created header. Fields of such header are initialized from header of input packet(except length field which value is depends on number of bytes copied to rx buffer). But in SEQPACKET case, we also need to take care of record delimeter bit: if input packet has this bit set, we don't copy it to header of packet in rx buffer, except case when such rx buffer is last part of input packet. Otherwise, we will get sequence of packets with delimeter bit set, thus braking record bounds. Also remove ignore of non-stream type of packets, handle SEQPACKET feature bit. Signed-off-by:
-
Arseny Krasnov authored
To make transport work with SOCK_SEQPACKET add two things: 1) SOCK_SEQPACKET ops for virtio transport and 'seqpacket_allow()' callback. 2) Handling of SEQPACKET bit: guest tries to negotiate it with vhost, so feature will be enabled only if bit is negotiated with device. Signed-off-by:
-
Arseny Krasnov authored
Small updates to make SOCK_SEQPACKET work: 1) Send SHUTDOWN on socket close for SEQPACKET type. 2) Set SEQPACKET packet type during send. 3) Set 'VIRTIO_VSOCK_SEQ_EOR' bit in flags for last packet of message. 4) Implement data check function for SEQPACKET. 5) Check for max datagram size. Signed-off-by:
-
Arseny Krasnov authored
Update current receive logic for SEQPACKET support: performs check for packet and socket types on receive(if mismatch, then reset connection). Increment EOR counter on receive. Also if buffer of new packet was appended to buffer of last packet in rx queue, update flags of last packet with flags of new packet. Signed-off-by:
-
Arseny Krasnov authored
Callback fetches RW packets from rx queue of socket until whole record is copied(if user's buffer is full, user is not woken up). This is done to not stall sender, because if we wake up user and it leaves syscall, nobody will send credit update for rest of record, and sender will wait for next enter of read syscall at receiver's side. So if user buffer is full, we just send credit update and drop data. Signed-off-by:
-
Arseny Krasnov authored
Add set of defines and constants for SOCK_SEQPACKET support in vsock. Signed-off-by:
Stefano Garzarella <sgarzare@redhat.com> Signed-off-by:
-
Arseny Krasnov authored
This function is static and 'hdr' arg was always NULL. Signed-off-by:
-
Arseny Krasnov authored
There is no need to set type of packet which differs from type of socket, so move passing type of packet from 'info' structure to 'virtio_transport_send_pkt_info()' function. Since at current time only stream type is supported, set it directly in 'virtio_ transport_send_pkt_info()', so callers don't need to set it. Signed-off-by:
-
Arseny Krasnov authored
Replace 'stream' to 'connection oriented' in comments as SEQPACKET is also connection oriented. Signed-off-by:
-
Arseny Krasnov authored
Add socket ops for SEQPACKET type and .seqpacket_allow() callback to query transports if they support SEQPACKET. Also split path for data check for STREAM and SEQPACKET branches. Signed-off-by:
-
