linux-stable.git/drivers/tty/serial, branch v5.4.161 Linux kernel stable tree serial: xilinx_uartps: Fix race condition causing stuck TX 2021-11-17T08:48:43+00:00 Anssi Hannula anssi.hannula@bitwise.fi 2021-10-26T10:27:41+00:00 fae0552d83ff44fd225ae3de1297d6075f1c652b [ Upstream commit 88b20f84f0fe47409342669caf3e58a3fc64c316 ] xilinx_uartps .start_tx() clears TXEMPTY when enabling TXEMPTY to avoid any previous TXEVENT event asserting the UART interrupt. This clear operation is done immediately after filling the TX FIFO. However, if the bytes inserted by cdns_uart_handle_tx() are consumed by the UART before the TXEMPTY is cleared, the clear operation eats the new TXEMPTY event as well, causing cdns_uart_isr() to never receive the TXEMPTY event. If there are bytes still queued in circbuf, TX will get stuck as they will never get transferred to FIFO (unless new bytes are queued to circbuf in which case .start_tx() is called again). While the racy missed TXEMPTY occurs fairly often with short data sequences (e.g. write 1 byte), in those cases circbuf is usually empty so no action on TXEMPTY would have been needed anyway. On the other hand, longer data sequences make the race much more unlikely as UART takes longer to consume the TX FIFO. Therefore it is rare for this race to cause visible issues in general. Fix the race by clearing the TXEMPTY bit in ISR *before* filling the FIFO. The TXEMPTY bit in ISR will only get asserted at the exact moment the TX FIFO *becomes* empty, so clearing the bit before filling FIFO does not cause an extra immediate assertion even if the FIFO is initially empty. This is hard to reproduce directly on a normal system, but inserting e.g. udelay(200) after cdns_uart_handle_tx(port), setting 4000000 baud, and then running "dd if=/dev/zero bs=128 of=/dev/ttyPS0 count=50" reliably reproduces the issue on my ZynqMP test system unless this fix is applied. Fixes: 85baf542d54e ("tty: xuartps: support 64 byte FIFO size") Signed-off-by: Anssi Hannula <anssi.hannula@bitwise.fi> Link: https://lore.kernel.org/r/20211026102741.2910441-1-anssi.hannula@bitwise.fi Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Sasha Levin <sashal@kernel.org> 
[ Upstream commit 88b20f84f0fe47409342669caf3e58a3fc64c316 ]

xilinx_uartps .start_tx() clears TXEMPTY when enabling TXEMPTY to avoid
any previous TXEVENT event asserting the UART interrupt. This clear
operation is done immediately after filling the TX FIFO.

However, if the bytes inserted by cdns_uart_handle_tx() are consumed by
the UART before the TXEMPTY is cleared, the clear operation eats the new
TXEMPTY event as well, causing cdns_uart_isr() to never receive the
TXEMPTY event. If there are bytes still queued in circbuf, TX will get
stuck as they will never get transferred to FIFO (unless new bytes are
queued to circbuf in which case .start_tx() is called again).

While the racy missed TXEMPTY occurs fairly often with short data
sequences (e.g. write 1 byte), in those cases circbuf is usually empty
so no action on TXEMPTY would have been needed anyway. On the other
hand, longer data sequences make the race much more unlikely as UART
takes longer to consume the TX FIFO. Therefore it is rare for this race
to cause visible issues in general.

Fix the race by clearing the TXEMPTY bit in ISR *before* filling the
FIFO.

The TXEMPTY bit in ISR will only get asserted at the exact moment the
TX FIFO *becomes* empty, so clearing the bit before filling FIFO does
not cause an extra immediate assertion even if the FIFO is initially
empty.

This is hard to reproduce directly on a normal system, but inserting
e.g. udelay(200) after cdns_uart_handle_tx(port), setting 4000000 baud,
and then running "dd if=/dev/zero bs=128 of=/dev/ttyPS0 count=50"
reliably reproduces the issue on my ZynqMP test system unless this fix
is applied.

Fixes: 85baf542d54e ("tty: xuartps: support 64 byte FIFO size")
Signed-off-by: Anssi Hannula <anssi.hannula@bitwise.fi>
Link: https://lore.kernel.org/r/20211026102741.2910441-1-anssi.hannula@bitwise.fi
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
serial: 8250_dw: Drop wrong use of ACPI_PTR() 2021-11-17T08:48:42+00:00 Andy Shevchenko andriy.shevchenko@linux.intel.com 2021-10-05T13:45:16+00:00 15355466cdedcbf60efa301f2fc89277b1318990 [ Upstream commit ebabb77a2a115b6c5e68f7364b598310b5f61fb2 ] ACPI_PTR() is more harmful than helpful. For example, in this case if CONFIG_ACPI=n, the ID table left unused which is not what we want. Instead of adding ifdeffery here and there, drop ACPI_PTR(). Fixes: 6a7320c4669f ("serial: 8250_dw: Add ACPI 5.0 support") Reported-by: Daniel Palmer <daniel@0x0f.com> Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Link: https://lore.kernel.org/r/20211005134516.23218-1-andriy.shevchenko@linux.intel.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Sasha Levin <sashal@kernel.org> 
[ Upstream commit ebabb77a2a115b6c5e68f7364b598310b5f61fb2 ]

ACPI_PTR() is more harmful than helpful. For example, in this case
if CONFIG_ACPI=n, the ID table left unused which is not what we want.

Instead of adding ifdeffery here and there, drop ACPI_PTR().

Fixes: 6a7320c4669f ("serial: 8250_dw: Add ACPI 5.0 support")
Reported-by: Daniel Palmer <daniel@0x0f.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Link: https://lore.kernel.org/r/20211005134516.23218-1-andriy.shevchenko@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
serial: core: Fix initializing and restoring termios speed 2021-11-17T08:48:25+00:00 Pali Rohár pali@kernel.org 2021-10-02T13:09:00+00:00 e9adf72bdbd576a9d171ad933c7c2d7139a2444b commit 027b57170bf8bb6999a28e4a5f3d78bf1db0f90c upstream. Since commit edc6afc54968 ("tty: switch to ktermios and new framework") termios speed is no longer stored only in c_cflag member but also in new additional c_ispeed and c_ospeed members. If BOTHER flag is set in c_cflag then termios speed is stored only in these new members. Therefore to correctly restore termios speed it is required to store also ispeed and ospeed members, not only cflag member. In case only cflag member with BOTHER flag is restored then functions tty_termios_baud_rate() and tty_termios_input_baud_rate() returns baudrate stored in c_ospeed / c_ispeed member, which is zero as it was not restored too. If reported baudrate is invalid (e.g. zero) then serial core functions report fallback baudrate value 9600. So it means that in this case original baudrate is lost and kernel changes it to value 9600. Simple reproducer of this issue is to boot kernel with following command line argument: "console=ttyXXX,86400" (where ttyXXX is the device name). For speed 86400 there is no Bnnn constant and therefore kernel has to represent this speed via BOTHER c_cflag. Which means that speed is stored only in c_ospeed and c_ispeed members, not in c_cflag anymore. If bootloader correctly configures serial device to speed 86400 then kernel prints boot log to early console at speed speed 86400 without any issue. But after kernel starts initializing real console device ttyXXX then speed is changed to fallback value 9600 because information about speed was lost. This patch fixes above issue by storing and restoring also ispeed and ospeed members, which are required for BOTHER flag. Fixes: edc6afc54968 ("[PATCH] tty: switch to ktermios and new framework") Cc: stable@vger.kernel.org Signed-off-by: Pali Rohár <pali@kernel.org> Link: https://lore.kernel.org/r/20211002130900.9518-1-pali@kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> 
commit 027b57170bf8bb6999a28e4a5f3d78bf1db0f90c upstream.

Since commit edc6afc54968 ("tty: switch to ktermios and new framework")
termios speed is no longer stored only in c_cflag member but also in new
additional c_ispeed and c_ospeed members. If BOTHER flag is set in c_cflag
then termios speed is stored only in these new members.

Therefore to correctly restore termios speed it is required to store also
ispeed and ospeed members, not only cflag member.

In case only cflag member with BOTHER flag is restored then functions
tty_termios_baud_rate() and tty_termios_input_baud_rate() returns baudrate
stored in c_ospeed / c_ispeed member, which is zero as it was not restored
too. If reported baudrate is invalid (e.g. zero) then serial core functions
report fallback baudrate value 9600. So it means that in this case original
baudrate is lost and kernel changes it to value 9600.

Simple reproducer of this issue is to boot kernel with following command
line argument: "console=ttyXXX,86400" (where ttyXXX is the device name).
For speed 86400 there is no Bnnn constant and therefore kernel has to
represent this speed via BOTHER c_cflag. Which means that speed is stored
only in c_ospeed and c_ispeed members, not in c_cflag anymore.

If bootloader correctly configures serial device to speed 86400 then kernel
prints boot log to early console at speed speed 86400 without any issue.
But after kernel starts initializing real console device ttyXXX then speed
is changed to fallback value 9600 because information about speed was lost.

This patch fixes above issue by storing and restoring also ispeed and
ospeed members, which are required for BOTHER flag.

Fixes: edc6afc54968 ("[PATCH] tty: switch to ktermios and new framework")
Cc: stable@vger.kernel.org
Signed-off-by: Pali Rohár <pali@kernel.org>
Link: https://lore.kernel.org/r/20211002130900.9518-1-pali@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
serial: mvebu-uart: fix driver's tx_empty callback 2021-09-30T08:09:22+00:00 Pali Rohár pali@kernel.org 2021-09-11T13:20:17+00:00 494260e20ac25cc281671ce3cf05b2bd1c66f8a3 commit 74e1eb3b4a1ef2e564b4bdeb6e92afe844e900de upstream. Driver's tx_empty callback should signal when the transmit shift register is empty. So when the last character has been sent. STAT_TX_FIFO_EMP bit signals only that HW transmit FIFO is empty, which happens when the last byte is loaded into transmit shift register. STAT_TX_EMP bit signals when the both HW transmit FIFO and transmit shift register are empty. So replace STAT_TX_FIFO_EMP check by STAT_TX_EMP in mvebu_uart_tx_empty() callback function. Fixes: 30530791a7a0 ("serial: mvebu-uart: initial support for Armada-3700 serial port") Cc: stable <stable@vger.kernel.org> Signed-off-by: Pali Rohár <pali@kernel.org> Link: https://lore.kernel.org/r/20210911132017.25505-1-pali@kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> 
commit 74e1eb3b4a1ef2e564b4bdeb6e92afe844e900de upstream.

Driver's tx_empty callback should signal when the transmit shift register
is empty. So when the last character has been sent.

STAT_TX_FIFO_EMP bit signals only that HW transmit FIFO is empty, which
happens when the last byte is loaded into transmit shift register.

STAT_TX_EMP bit signals when the both HW transmit FIFO and transmit shift
register are empty.

So replace STAT_TX_FIFO_EMP check by STAT_TX_EMP in mvebu_uart_tx_empty()
callback function.

Fixes: 30530791a7a0 ("serial: mvebu-uart: initial support for Armada-3700 serial port")
Cc: stable <stable@vger.kernel.org>
Signed-off-by: Pali Rohár <pali@kernel.org>
Link: https://lore.kernel.org/r/20210911132017.25505-1-pali@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
serial: sh-sci: fix break handling for sysrq 2021-09-22T10:26:33+00:00 Ulrich Hecht uli+renesas@fpond.eu 2021-08-16T16:22:01+00:00 2918eca4970a1434cb3890f8ddfd39758823c97a [ Upstream commit 87b8061bad9bd4b549b2daf36ffbaa57be2789a2 ] This fixes two issues that cause the sysrq sequence to be inadvertently aborted on SCIF serial consoles: - a NUL character remains in the RX queue after a break has been detected, which is then passed on to uart_handle_sysrq_char() - the break interrupt is handled twice on controllers with multiplexed ERI and BRI interrupts Signed-off-by: Ulrich Hecht <uli+renesas@fpond.eu> Link: https://lore.kernel.org/r/20210816162201.28801-1-uli+renesas@fpond.eu Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Sasha Levin <sashal@kernel.org> 
[ Upstream commit 87b8061bad9bd4b549b2daf36ffbaa57be2789a2 ]

This fixes two issues that cause the sysrq sequence to be inadvertently
aborted on SCIF serial consoles:

- a NUL character remains in the RX queue after a break has been detected,
  which is then passed on to uart_handle_sysrq_char()
- the break interrupt is handled twice on controllers with multiplexed ERI
  and BRI interrupts

Signed-off-by: Ulrich Hecht <uli+renesas@fpond.eu>
Link: https://lore.kernel.org/r/20210816162201.28801-1-uli+renesas@fpond.eu
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
serial: 8250_pci: make setup_port() parameters explicitly unsigned 2021-09-22T10:26:30+00:00 Greg Kroah-Hartman gregkh@linuxfoundation.org 2021-07-26T13:07:17+00:00 44fd61a8bd0dc543a1bdde557f0cb4487ac44ffd [ Upstream commit 3a96e97ab4e835078e6f27b7e1c0947814df3841 ] The bar and offset parameters to setup_port() are used in pointer math, and while it would be very difficult to get them to wrap as a negative number, just be "safe" and make them unsigned so that static checkers do not trip over them unintentionally. Cc: Jiri Slaby <jirislaby@kernel.org> Reported-by: Jordy Zomer <jordy@pwning.systems> Link: https://lore.kernel.org/r/20210726130717.2052096-1-gregkh@linuxfoundation.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Sasha Levin <sashal@kernel.org> 
[ Upstream commit 3a96e97ab4e835078e6f27b7e1c0947814df3841 ]

The bar and offset parameters to setup_port() are used in pointer math,
and while it would be very difficult to get them to wrap as a negative
number, just be "safe" and make them unsigned so that static checkers do
not trip over them unintentionally.

Cc: Jiri Slaby <jirislaby@kernel.org>
Reported-by: Jordy Zomer <jordy@pwning.systems>
Link: https://lore.kernel.org/r/20210726130717.2052096-1-gregkh@linuxfoundation.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
serial: 8250: Define RX trigger levels for OxSemi 950 devices 2021-09-22T10:26:29+00:00 Maciej W. Rozycki macro@orcam.me.uk 2021-06-26T04:11:51+00:00 5418023f81cd84721a84e9fa8a83ae9abdd0b315 [ Upstream commit d7aff291d069c4418285f3c8ee27b0ff67ce5998 ] Oxford Semiconductor 950 serial port devices have a 128-byte FIFO and in the enhanced (650) mode, which we select in `autoconfig_has_efr' with the ECB bit set in the EFR register, they support the receive interrupt trigger level selectable with FCR bits 7:6 from the set of 16, 32, 112, 120. This applies to the original OX16C950 discrete UART[1] as well as 950 cores embedded into more complex devices. For these devices we set the default to 112, which sets an excessively high level of 112 or 7/8 of the FIFO capacity, unlike with other port types where we choose at most 1/2 of their respective FIFO capacities. Additionally we don't make the trigger level configurable. Consequently frequent input overruns happen with high bit rates where hardware flow control cannot be used (e.g. terminal applications) even with otherwise highly-performant systems. Lower the default receive interrupt trigger level to 32 then, and make it configurable. Document the trigger levels along with other port types, including the set of 16, 32, 64, 112 for the transmit interrupt as well[2]. References: [1] "OX16C950 rev B High Performance UART with 128 byte FIFOs", Oxford Semiconductor, Inc., DS-0031, Sep 05, Table 10: "Receiver Trigger Levels", p. 22 [2] same, Table 9: "Transmit Interrupt Trigger Levels", p. 22 Signed-off-by: Maciej W. Rozycki <macro@orcam.me.uk> Link: https://lore.kernel.org/r/alpine.DEB.2.21.2106260608480.37803@angie.orcam.me.uk Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Sasha Levin <sashal@kernel.org> 
[ Upstream commit d7aff291d069c4418285f3c8ee27b0ff67ce5998 ]

Oxford Semiconductor 950 serial port devices have a 128-byte FIFO and in
the enhanced (650) mode, which we select in `autoconfig_has_efr' with
the ECB bit set in the EFR register, they support the receive interrupt
trigger level selectable with FCR bits 7:6 from the set of 16, 32, 112,
120.  This applies to the original OX16C950 discrete UART[1] as well as
950 cores embedded into more complex devices.

For these devices we set the default to 112, which sets an excessively
high level of 112 or 7/8 of the FIFO capacity, unlike with other port
types where we choose at most 1/2 of their respective FIFO capacities.
Additionally we don't make the trigger level configurable.  Consequently
frequent input overruns happen with high bit rates where hardware flow
control cannot be used (e.g. terminal applications) even with otherwise
highly-performant systems.

Lower the default receive interrupt trigger level to 32 then, and make
it configurable.  Document the trigger levels along with other port
types, including the set of 16, 32, 64, 112 for the transmit interrupt
as well[2].

References:

[1] "OX16C950 rev B High Performance UART with 128 byte FIFOs", Oxford
    Semiconductor, Inc., DS-0031, Sep 05, Table 10: "Receiver Trigger
    Levels", p. 22

[2] same, Table 9: "Transmit Interrupt Trigger Levels", p. 22

Signed-off-by: Maciej W. Rozycki <macro@orcam.me.uk>
Link: https://lore.kernel.org/r/alpine.DEB.2.21.2106260608480.37803@angie.orcam.me.uk
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
tty: serial: jsm: hold port lock when reporting modem line changes 2021-09-22T10:26:28+00:00 Zheyu Ma zheyuma97@gmail.com 2021-07-14T05:53:23+00:00 6c78ee1aecb927e5e372fb254bd00563e1675ea0 [ Upstream commit 240e126c28df084222f0b661321e8e3ecb0d232e ] uart_handle_dcd_change() requires a port lock to be held and will emit a warning when lockdep is enabled. Held corresponding lock to fix the following warnings. [ 132.528648] WARNING: CPU: 5 PID: 11600 at drivers/tty/serial/serial_core.c:3046 uart_handle_dcd_change+0xf4/0x120 [ 132.530482] Modules linked in: [ 132.531050] CPU: 5 PID: 11600 Comm: jsm Not tainted 5.14.0-rc1-00003-g7fef2edf7cc7-dirty #31 [ 132.535268] RIP: 0010:uart_handle_dcd_change+0xf4/0x120 [ 132.557100] Call Trace: [ 132.557562] ? __free_pages+0x83/0xb0 [ 132.558213] neo_parse_modem+0x156/0x220 [ 132.558897] neo_param+0x399/0x840 [ 132.559495] jsm_tty_open+0x12f/0x2d0 [ 132.560131] uart_startup.part.18+0x153/0x340 [ 132.560888] ? lock_is_held_type+0xe9/0x140 [ 132.561660] uart_port_activate+0x7f/0xe0 [ 132.562351] ? uart_startup.part.18+0x340/0x340 [ 132.563003] tty_port_open+0x8d/0xf0 [ 132.563523] ? uart_set_options+0x1e0/0x1e0 [ 132.564125] uart_open+0x24/0x40 [ 132.564604] tty_open+0x15c/0x630 Signed-off-by: Zheyu Ma <zheyuma97@gmail.com> Link: https://lore.kernel.org/r/1626242003-3809-1-git-send-email-zheyuma97@gmail.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Sasha Levin <sashal@kernel.org> 
[ Upstream commit 240e126c28df084222f0b661321e8e3ecb0d232e ]

uart_handle_dcd_change() requires a port lock to be held and will emit a
warning when lockdep is enabled.

Held corresponding lock to fix the following warnings.

[  132.528648] WARNING: CPU: 5 PID: 11600 at drivers/tty/serial/serial_core.c:3046 uart_handle_dcd_change+0xf4/0x120
[  132.530482] Modules linked in:
[  132.531050] CPU: 5 PID: 11600 Comm: jsm Not tainted 5.14.0-rc1-00003-g7fef2edf7cc7-dirty #31
[  132.535268] RIP: 0010:uart_handle_dcd_change+0xf4/0x120
[  132.557100] Call Trace:
[  132.557562]  ? __free_pages+0x83/0xb0
[  132.558213]  neo_parse_modem+0x156/0x220
[  132.558897]  neo_param+0x399/0x840
[  132.559495]  jsm_tty_open+0x12f/0x2d0
[  132.560131]  uart_startup.part.18+0x153/0x340
[  132.560888]  ? lock_is_held_type+0xe9/0x140
[  132.561660]  uart_port_activate+0x7f/0xe0
[  132.562351]  ? uart_startup.part.18+0x340/0x340
[  132.563003]  tty_port_open+0x8d/0xf0
[  132.563523]  ? uart_set_options+0x1e0/0x1e0
[  132.564125]  uart_open+0x24/0x40
[  132.564604]  tty_open+0x15c/0x630

Signed-off-by: Zheyu Ma <zheyuma97@gmail.com>
Link: https://lore.kernel.org/r/1626242003-3809-1-git-send-email-zheyuma97@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
tty: serial: fsl_lpuart: fix the wrong mapbase value 2021-09-15T07:47:37+00:00 Andy Duan fugang.duan@nxp.com 2021-08-19T02:10:33+00:00 7bfa680f3b478e699ba96f9809384502d6f386e7 [ Upstream commit d5c38948448abc2bb6b36dbf85a554bf4748885e ] Register offset needs to be applied on mapbase also. dma_tx/rx_request use the physical address of UARTDATA. Register offset is currently only applied to membase (the corresponding virtual addr) but not on mapbase. Fixes: 24b1e5f0e83c ("tty: serial: lpuart: add imx7ulp support") Reviewed-by: Leonard Crestez <leonard.crestez@nxp.com> Signed-off-by: Adriana Reus <adriana.reus@nxp.com> Signed-off-by: Sherry Sun <sherry.sun@nxp.com> Signed-off-by: Andy Duan <fugang.duan@nxp.com> Link: https://lore.kernel.org/r/20210819021033.32606-1-sherry.sun@nxp.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Sasha Levin <sashal@kernel.org> 
[ Upstream commit d5c38948448abc2bb6b36dbf85a554bf4748885e ]

Register offset needs to be applied on mapbase also.
dma_tx/rx_request use the physical address of UARTDATA.
Register offset is currently only applied to membase (the
corresponding virtual addr) but not on mapbase.

Fixes: 24b1e5f0e83c ("tty: serial: lpuart: add imx7ulp support")
Reviewed-by: Leonard Crestez <leonard.crestez@nxp.com>
Signed-off-by: Adriana Reus <adriana.reus@nxp.com>
Signed-off-by: Sherry Sun <sherry.sun@nxp.com>
Signed-off-by: Andy Duan <fugang.duan@nxp.com>
Link: https://lore.kernel.org/r/20210819021033.32606-1-sherry.sun@nxp.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
serial: 8250_pci: Avoid irq sharing for MSI(-X) interrupts. 2021-08-12T11:21:03+00:00 Mario Kleiner mario.kleiner.de@gmail.com 2021-07-29T04:33:06+00:00 8211bb20da2352cd324caa44aaf371e81b2e7ea2 commit 341abd693d10e5f337a51f140ae3e7a1ae0febf6 upstream. This attempts to fix a bug found with a serial port card which uses an MCS9922 chip, one of the 4 models for which MSI-X interrupts are currently supported. I don't possess such a card, and i'm not experienced with the serial subsystem, so this patch is based on what i think i found as a likely reason for failure, based on walking the user who actually owns the card through some diagnostic. The user who reported the problem finds the following in his dmesg output for the relevant ttyS4 and ttyS5: [ 0.580425] serial 0000:02:00.0: enabling device (0000 -> 0003) [ 0.601448] 0000:02:00.0: ttyS4 at I/O 0x3010 (irq = 125, base_baud = 115200) is a ST16650V2 [ 0.603089] serial 0000:02:00.1: enabling device (0000 -> 0003) [ 0.624119] 0000:02:00.1: ttyS5 at I/O 0x3000 (irq = 126, base_baud = 115200) is a ST16650V2 ... [ 6.323784] genirq: Flags mismatch irq 128. 00000080 (ttyS5) vs. 00000000 (xhci_hcd) [ 6.324128] genirq: Flags mismatch irq 128. 00000080 (ttyS5) vs. 00000000 (xhci_hcd) ... Output of setserial -a: /dev/ttyS4, Line 4, UART: 16650V2, Port: 0x3010, IRQ: 127 Baud_base: 115200, close_delay: 50, divisor: 0 closing_wait: 3000 Flags: spd_normal skip_test This suggests to me that the serial driver wants to register and share a MSI/MSI-X irq 128 with the xhci_hcd driver, whereas the xhci driver does not want to share the irq, as flags 0x00000080 (== IRQF_SHARED) from the serial port driver means to share the irq, and this mismatch ends in some failed irq init? With this setup, data reception works very unreliable, with dropped data, already at a transmission rate of only a 16 Bytes chunk every 1/120th of a second, ie. 1920 Bytes/sec, presumably due to rx fifo overflow due to mishandled or not used at all rx irq's? See full discussion thread with attempted diagnosis at: https://psychtoolbox.discourse.group/t/issues-with-iscan-serial-port-recording/3886 Disabling the use of MSI interrupts for the serial port pci card did fix the reliability problems. The user executed the following sequence of commands to achieve this: echo 0000:02:00.0 | sudo tee /sys/bus/pci/drivers/serial/unbind echo 0000:02:00.1 | sudo tee /sys/bus/pci/drivers/serial/unbind echo 0 | sudo tee /sys/bus/pci/devices/0000:02:00.0/msi_bus echo 0 | sudo tee /sys/bus/pci/devices/0000:02:00.1/msi_bus echo 0000:02:00.0 | sudo tee /sys/bus/pci/drivers/serial/bind echo 0000:02:00.1 | sudo tee /sys/bus/pci/drivers/serial/bind This resulted in the following log output: [ 82.179021] pci 0000:02:00.0: MSI/MSI-X disallowed for future drivers [ 87.003031] pci 0000:02:00.1: MSI/MSI-X disallowed for future drivers [ 98.537010] 0000:02:00.0: ttyS4 at I/O 0x3010 (irq = 17, base_baud = 115200) is a ST16650V2 [ 103.648124] 0000:02:00.1: ttyS5 at I/O 0x3000 (irq = 18, base_baud = 115200) is a ST16650V2 This patch attempts to fix the problem by disabling irq sharing when using MSI irq's. Note that all i know for sure is that disabling MSI irq's fixed the problem for the user, so this patch could be wrong and is untested. Please review with caution, keeping this in mind. Fixes: 8428413b1d14 ("serial: 8250_pci: Implement MSI(-X) support") Cc: Ralf Ramsauer <ralf.ramsauer@oth-regensburg.de> Cc: stable <stable@vger.kernel.org> Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com> Signed-off-by: Mario Kleiner <mario.kleiner.de@gmail.com> Link: https://lore.kernel.org/r/20210729043306.18528-1-mario.kleiner.de@gmail.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> 
commit 341abd693d10e5f337a51f140ae3e7a1ae0febf6 upstream.

This attempts to fix a bug found with a serial port card which uses
an MCS9922 chip, one of the 4 models for which MSI-X interrupts are
currently supported. I don't possess such a card, and i'm not
experienced with the serial subsystem, so this patch is based on what
i think i found as a likely reason for failure, based on walking the
user who actually owns the card through some diagnostic.

The user who reported the problem finds the following in his dmesg
output for the relevant ttyS4 and ttyS5:

[    0.580425] serial 0000:02:00.0: enabling device (0000 -> 0003)
[    0.601448] 0000:02:00.0: ttyS4 at I/O 0x3010 (irq = 125, base_baud = 115200) is a ST16650V2
[    0.603089] serial 0000:02:00.1: enabling device (0000 -> 0003)
[    0.624119] 0000:02:00.1: ttyS5 at I/O 0x3000 (irq = 126, base_baud = 115200) is a ST16650V2
...
[    6.323784] genirq: Flags mismatch irq 128. 00000080 (ttyS5) vs. 00000000 (xhci_hcd)
[    6.324128] genirq: Flags mismatch irq 128. 00000080 (ttyS5) vs. 00000000 (xhci_hcd)
...

Output of setserial -a:

/dev/ttyS4, Line 4, UART: 16650V2, Port: 0x3010, IRQ: 127
	Baud_base: 115200, close_delay: 50, divisor: 0
	closing_wait: 3000
	Flags: spd_normal skip_test

This suggests to me that the serial driver wants to register and share a
MSI/MSI-X irq 128 with the xhci_hcd driver, whereas the xhci driver does
not want to share the irq, as flags 0x00000080 (== IRQF_SHARED) from the
serial port driver means to share the irq, and this mismatch ends in some
failed irq init?

With this setup, data reception works very unreliable, with dropped data,
already at a transmission rate of only a 16 Bytes chunk every 1/120th of
a second, ie. 1920 Bytes/sec, presumably due to rx fifo overflow due to
mishandled or not used at all rx irq's?

See full discussion thread with attempted diagnosis at:

https://psychtoolbox.discourse.group/t/issues-with-iscan-serial-port-recording/3886

Disabling the use of MSI interrupts for the serial port pci card did
fix the reliability problems. The user executed the following sequence
of commands to achieve this:

echo 0000:02:00.0 | sudo tee /sys/bus/pci/drivers/serial/unbind
echo 0000:02:00.1 | sudo tee /sys/bus/pci/drivers/serial/unbind

echo 0 | sudo tee /sys/bus/pci/devices/0000:02:00.0/msi_bus
echo 0 | sudo tee /sys/bus/pci/devices/0000:02:00.1/msi_bus

echo 0000:02:00.0 | sudo tee /sys/bus/pci/drivers/serial/bind
echo 0000:02:00.1 | sudo tee /sys/bus/pci/drivers/serial/bind

This resulted in the following log output:

[   82.179021] pci 0000:02:00.0: MSI/MSI-X disallowed for future drivers
[   87.003031] pci 0000:02:00.1: MSI/MSI-X disallowed for future drivers
[   98.537010] 0000:02:00.0: ttyS4 at I/O 0x3010 (irq = 17, base_baud = 115200) is a ST16650V2
[  103.648124] 0000:02:00.1: ttyS5 at I/O 0x3000 (irq = 18, base_baud = 115200) is a ST16650V2

This patch attempts to fix the problem by disabling irq sharing when
using MSI irq's. Note that all i know for sure is that disabling MSI
irq's fixed the problem for the user, so this patch could be wrong and
is untested. Please review with caution, keeping this in mind.

Fixes: 8428413b1d14 ("serial: 8250_pci: Implement MSI(-X) support")
Cc: Ralf Ramsauer <ralf.ramsauer@oth-regensburg.de>
Cc: stable <stable@vger.kernel.org>
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Signed-off-by: Mario Kleiner <mario.kleiner.de@gmail.com>
Link: https://lore.kernel.org/r/20210729043306.18528-1-mario.kleiner.de@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>