Page MenuHomeFreeBSD

hayzam (Hayzam Sherif)
User

Projects

User is not a member of any projects.

User Details

User Since
Aug 24 2025, 12:57 PM (64 w, 2 d)

Recent Activity

Today

hayzam committed rG5b2f5e9999a9: bhyve: Do not abort when the VM is destroyed during an exit (authored by wanpengqian_gmail.com).
Tue, Nov 17, 7:04 PM
hayzam closed D61376: bhyve: Do not abort when the VM is destroyed during an exit.
Tue, Nov 17, 7:04 PM
hayzam added inline comments to D61266: bhyve: fix snapshot IPC handling.
Tue, Nov 17, 4:57 PM
hayzam added inline comments to D61191: uart: Add MediaTek UART driver.
Tue, Nov 17, 4:38 PM
hayzam updated the diff for D61191: uart: Add MediaTek UART driver.

Reword the five data bit note in the header comment

Tue, Nov 17, 4:34 PM
hayzam accepted D61376: bhyve: Do not abort when the VM is destroyed during an exit.

LGTM, but I'll wait for others to look it over as well.

Tue, Nov 17, 1:41 PM
pi renamed hayzam.
Tue, Nov 17, 1:34 PM
hayzam added a comment to D61376: bhyve: Do not abort when the VM is destroyed during an exit.

Thanks for testing this and for both comments.

  1. inout.c: agreed, that is better. The diff now keeps the asserts and calls vm_check_destroyed() right before each of them, so on a VM that still exists the assert names the failing call again, and "Unhandled in/out" is only printed for a handler error.
  1. Scope: not intended, I had only covered the paths I had hit. vm_check_destroyed() moved to bhyverun.c (it probes with vm_get_capability(VM_CAP_UNRESTRICTED_GUEST), which all three architectures have) and is now also called before the asserts in vmexit_rdmsr() and vm_inject_fault(), and on the vmexit_inst_emul() failure paths on arm64 and riscv. With an rdmsr 0x17 loop (through cpuctl, one process per vCPU) I get 8/10 SIGABRT in vmexit_rdmsr() without the change and with the first version, and 10/10 exit 4 with this one; the port I/O load gives 4/10 SIGABRT vs. 10/10 exit 4. Details are in the test plan.

I have no arm64 or riscv machine to run bhyve on, so those two hunks are only syntax-checked with cc --target. Would you prefer to keep them here or move them to a separate review? markj, bnovkov: the same question to you.

This touches the same lines in inout.c as D61498 and D61499; whichever lands second needs a trivial rebase.

Tue, Nov 17, 1:08 PM
hayzam added a reviewer for D61376: bhyve: Do not abort when the VM is destroyed during an exit: hayzam.
Tue, Nov 17, 6:50 AM
hayzam added inline comments to D61376: bhyve: Do not abort when the VM is destroyed during an exit.
Tue, Nov 17, 6:50 AM
hayzam added a comment to D61426: bhyve: register a BAR again once the range it wanted is free.

Thanks. I tested this on amd64 main (7cb4d4dfdd18) with a FreeBSD main guest and two e1000 devices whose drivers were detached. The reproducer from the test plan reads ffffffff without the change and the moved BAR's registers with it. Without the change, one more write after that sequence, pciconf -w pci0:0:9:0 0x10 0xc0040000 to move slot 9's BAR back, also aborts bhyve on assert(error == 0) in modify_bar_registration(), because that BAR was never really registered at 0xc0020000. This change fixes that too, which may be worth saying in the commit message. FreeBSD and Alpine Linux guests with virtio, e1000, ahci, xhci and fbuf devices boot as before.

Tue, Nov 17, 6:07 AM
hayzam added a comment to D61376: bhyve: Do not abort when the VM is destroyed during an exit.
Tue, Nov 17, 6:06 AM
hayzam added a reviewer for D61426: bhyve: register a BAR again once the range it wanted is free: hayzam.
Tue, Nov 17, 5:55 AM
hayzam added inline comments to D61426: bhyve: register a BAR again once the range it wanted is free.
Tue, Nov 17, 5:54 AM
hayzam committed rG4ad4ddca0818: bhyve: Reject unsupported checkpoints before saving RAM.
Tue, Nov 17, 12:42 AM
hayzam committed rGee9e945e7c59: committers-src: Add hayzam mentored by markj@ and emaste@.
Tue, Nov 17, 12:42 AM
hayzam closed D61003: bhyve: Reject unsupported checkpoints before saving RAM.
Tue, Nov 17, 12:42 AM
hayzam committed R9:51ff07751d6d: Add new src committer (hayzam).
Tue, Nov 17, 12:33 AM

Yesterday

hayzam updated the summary of D61191: uart: Add MediaTek UART driver.
Mon, Nov 16, 11:14 PM
hayzam updated the diff for D61191: uart: Add MediaTek UART driver.

Fold the frame helpers, require 8 samples per bit as the manual says, fix uc_range units, small cleanups

Mon, Nov 16, 11:12 PM
hayzam updated the summary of D61191: uart: Add MediaTek UART driver.
Mon, Nov 16, 1:52 PM
hayzam updated the diff for D61191: uart: Add MediaTek UART driver.

Use the ns8250 baud rate generator hooks (D61611); the driver now only provides the generator

Mon, Nov 16, 1:49 PM
hayzam updated the test plan for D61622: arm64/pmap: Ignore pmap_advise() for stage 2 pmaps.
Mon, Nov 16, 1:38 PM
hayzam requested review of D61622: arm64/pmap: Ignore pmap_advise() for stage 2 pmaps.
Mon, Nov 16, 1:37 PM
hayzam added a comment to D61611: uart: Let ns8250 drivers provide their own baud rate generator.
In D61611#1411961, @hayzam_gmail.com wrote:

Yes, that's better than the extra argument. One question: ns8250_get_divisor()
only gets the bas, so outside probe should it reach the hooks with __containerof()
on the softc, or would you rather keep a pointer in the bas (e.g. the existing driver1 field)?

imho, you could change the first argument of the affected functions from 'struct uart_base *bas' to 'struct ns8250_softc *ns8250'. if not, __containerof() is also OK

Mon, Nov 16, 1:12 PM
hayzam updated the summary of D61611: uart: Let ns8250 drivers provide their own baud rate generator.
Mon, Nov 16, 1:10 PM
hayzam updated the diff for D61611: uart: Let ns8250 drivers provide their own baud rate generator.

Pass the softc to the helpers and use UART_BAS_IN_PROBE instead of passing get_divisor

Mon, Nov 16, 1:09 PM
hayzam requested review of D61620: uart: Add a flags field to struct uart_bas.
Mon, Nov 16, 1:09 PM
hayzam updated the summary of D61610: uart: Factor out the ns8250 baud rate generator code.
Mon, Nov 16, 12:36 PM
hayzam added inline comments to D61610: uart: Factor out the ns8250 baud rate generator code.
Mon, Nov 16, 12:28 PM
hayzam updated the diff for D61611: uart: Let ns8250 drivers provide their own baud rate generator.

Restore the original order in ns8250_param()

Mon, Nov 16, 12:26 PM
hayzam updated the diff for D61610: uart: Factor out the ns8250 baud rate generator code.

Restore the original order in ns8250_param()

Mon, Nov 16, 12:25 PM
hayzam added inline comments to D61610: uart: Factor out the ns8250 baud rate generator code.
Mon, Nov 16, 12:20 PM
hayzam updated the summary of D61610: uart: Factor out the ns8250 baud rate generator code.
Mon, Nov 16, 12:19 PM
hayzam added a comment to D61611: uart: Let ns8250 drivers provide their own baud rate generator.

Instead of adding get_divisor() to lots of functions, can we use the following approach?

  • Add a 'flags' field to the 'struct uart_bas' and define 'UART_BAS_IN_PROBE' (or any better name).
  • Set this flag as appropriate in the probe function.
  • Use this flag in ns8250_get_divisor() to select the correct function.

Is this OK with you?
if yes, it should be (probably) included in own review or in D61610.

Mon, Nov 16, 12:08 PM
hayzam updated the summary of D61610: uart: Factor out the ns8250 baud rate generator code.
Mon, Nov 16, 10:26 AM
hayzam updated the diff for D61611: uart: Let ns8250 drivers provide their own baud rate generator.

set_baudrate() no longer takes the LCR value: ns8250_param() writes LCR first and restores it if the rate is rejected

Mon, Nov 16, 10:21 AM
hayzam updated the diff for D61610: uart: Factor out the ns8250 baud rate generator code.

set_baudrate() no longer takes the LCR value: ns8250_param() writes LCR first and restores it if the rate is rejected

Mon, Nov 16, 10:20 AM
hayzam requested review of D61611: uart: Let ns8250 drivers provide their own baud rate generator.
Mon, Nov 16, 10:04 AM
hayzam requested review of D61610: uart: Factor out the ns8250 baud rate generator code.
Mon, Nov 16, 10:04 AM

Sun, Nov 15

hayzam updated the summary of D61186: arm64: Add MediaTek SoC support skeleton.
Sun, Nov 15, 9:22 PM
hayzam updated the diff for D61186: arm64: Add MediaTek SoC support skeleton.

Add a MEDIATEK kernel configuration, as the other SoC families have.

Sun, Nov 15, 9:21 PM
hayzam updated the summary of D61189: clk/mediatek: Add MT7986 TOPCKGEN clock controller.
Sun, Nov 15, 8:23 PM
hayzam updated the summary of D61186: arm64: Add MediaTek SoC support skeleton.
Sun, Nov 15, 8:17 PM
hayzam updated the diff for D61186: arm64: Add MediaTek SoC support skeleton.

List the SoC options in NOTES so that LINT builds the drivers.

Sun, Nov 15, 8:16 PM
hayzam updated the summary of D61190: clk/mediatek: Add MT7986 INFRACFG clock and reset controller.
Sun, Nov 15, 7:55 PM
hayzam updated the summary of D61189: clk/mediatek: Add MT7986 TOPCKGEN clock controller.
Sun, Nov 15, 7:54 PM
hayzam updated the summary of D61187: clk/mediatek: Add common MediaTek clock and reset support.
Sun, Nov 15, 7:52 PM
hayzam updated the diff for D61190: clk/mediatek: Add MT7986 INFRACFG clock and reset controller.

Muxes carry MTK_MUX_SETCLR: the status registers are read-only (checked on the board). Reset id list dropped.

Sun, Nov 15, 7:49 PM
hayzam updated the diff for D61189: clk/mediatek: Add MT7986 TOPCKGEN clock controller.

Muxes change by read-modify-write plus the update register. top_net2pll_d3_d2 divides by 6 (measured with the SoC's frequency meter).

Sun, Nov 15, 7:49 PM
hayzam updated the diff for D61187: clk/mediatek: Add common MediaTek clock and reset support.

Muxes change by read-modify-write unless the register is read-only. PLL set_freq sets the rate. Critical clocks use clknode_enable(). Reset ids are bounds-checked instead of listed.

Sun, Nov 15, 7:48 PM
hayzam updated the summary of D61191: uart: Add MediaTek UART driver.
Sun, Nov 15, 5:05 PM
hayzam updated the diff for D61191: uart: Add MediaTek UART driver.

Clocks moved to UART_ATTACH, softc via class size, as discussed

Sun, Nov 15, 5:03 PM
hayzam added a comment to D61191: uart: Add MediaTek UART driver.

Sorry, but the whole code relating to clocks is wrong. Why can't you use the standard functions provided by the clk framework? I don't want to continue until we clarify this area.

To put your mind at ease, I can't see any major issues with the rest of the code at first glance :)

Sun, Nov 15, 3:57 PM
hayzam added a reviewer for D61187: clk/mediatek: Add common MediaTek clock and reset support: manu.
Sun, Nov 15, 12:12 PM
hayzam added a reviewer for D61188: clk/mediatek: Add MT7986 APMIXEDSYS PLL controller: manu.
Sun, Nov 15, 12:12 PM
hayzam added a reviewer for D61189: clk/mediatek: Add MT7986 TOPCKGEN clock controller: manu.
Sun, Nov 15, 12:12 PM
hayzam added a reviewer for D61190: clk/mediatek: Add MT7986 INFRACFG clock and reset controller: manu.
Sun, Nov 15, 12:12 PM
hayzam added a reviewer for D61533: sysutils/fiptool: Add new port: manu.
Sun, Nov 15, 11:56 AM
hayzam added inline comments to D61539: sysutils/u-boot-bananapi-r3: Add new port.
Sun, Nov 15, 11:55 AM
hayzam added a comment to D61539: sysutils/u-boot-bananapi-r3: Add new port.

I really don't like having patches for u-boot, where those submitted upstream ?

Sun, Nov 15, 11:37 AM
hayzam added inline comments to D61539: sysutils/u-boot-bananapi-r3: Add new port.
Sun, Nov 15, 11:33 AM
hayzam added a reviewer for D61536: sysutils/atf-mt7986: Add new port: manu.
Sun, Nov 15, 11:31 AM

Sat, Nov 14

hayzam added a comment to D61536: sysutils/atf-mt7986: Add new port.
In D61536#1411154, @jrm wrote:

The commit message and pkg-descr say the DRAM initialization is shipped as a prebuilt library. From the build log, it looks like other prebuilt archives are on the link lines as well:

  • BL2: plat/mediatek/mt7986/drivers/dram/release/dram.a
  • BL2 and BL31: plat/mediatek/mt7986/drivers/efuse/release/plat_efuse.a and plat/mediatek/mt7986/drivers/efuse/release/efuse_cmd.a
  • BL2 and BL31: plat/mediatek/apsoc_common/img_dec/roe/release/mtk_plat_key.a, though this one may not be used in this configuration, since its only caller is built when FW_ENC or FIP_ENC is set.

I haven't checked which of these end up in the final images. Do you know what licence these are distributed under? I couldn't find any licence text for them in the tarball, and docs/license.rst doesn't mention them, so I'm not sure LICENSE=BSD3CLAUSE covers the whole package.

Sat, Nov 14, 9:48 PM
hayzam added inline comments to D61533: sysutils/fiptool: Add new port.
Sat, Nov 14, 9:25 PM
hayzam updated the diff for D61533: sysutils/fiptool: Add new port.

Fix fiptool version (FIPTOOL_CPPFLAGS, BUILD_STRING); drop -Werror

Sat, Nov 14, 9:23 PM
hayzam updated the diff for D61539: sysutils/u-boot-bananapi-r3: Add new port.

Drop CONFIG_DEFAULT_FDT_FILE from the fragment; it has no effect on MT7986

Sat, Nov 14, 11:13 AM
hayzam added a reviewer for D61540: sysutils/u-boot-bananapi-r3-emmc: Add new port: manu.
Sat, Nov 14, 10:48 AM
hayzam added a reviewer for D61539: sysutils/u-boot-bananapi-r3: Add new port: manu.
Sat, Nov 14, 10:48 AM
hayzam updated the test plan for D61540: sysutils/u-boot-bananapi-r3-emmc: Add new port.
Sat, Nov 14, 10:45 AM
hayzam updated the test plan for D61539: sysutils/u-boot-bananapi-r3: Add new port.
Sat, Nov 14, 10:45 AM
hayzam requested review of D61540: sysutils/u-boot-bananapi-r3-emmc: Add new port.
Sat, Nov 14, 10:45 AM
hayzam requested review of D61539: sysutils/u-boot-bananapi-r3: Add new port.
Sat, Nov 14, 10:45 AM
hayzam updated the test plan for D61536: sysutils/atf-mt7986: Add new port.
Sat, Nov 14, 8:56 AM
hayzam requested review of D61536: sysutils/atf-mt7986: Add new port.
Sat, Nov 14, 8:55 AM
hayzam updated the summary of D61533: sysutils/fiptool: Add new port.
Sat, Nov 14, 7:03 AM
hayzam published D61533: sysutils/fiptool: Add new port for review.
Sat, Nov 14, 7:02 AM
hayzam requested review of D61530: libvmmapi: Compare mapping lengths in vm_mmap_memseg().
Sat, Nov 14, 3:57 AM

Fri, Nov 13

hayzam added a comment to D61498: bhyve: Preserve upper RAX bits for byte and word input.

Does emulate_inout_port() in the kernel need the same change?

Fri, Nov 13, 5:13 PM
hayzam updated the diff for D61498: bhyve: Preserve upper RAX bits for byte and word input.

Fix RAX preservation in the kernel scalar IN handler too.

Fri, Nov 13, 5:13 PM
hayzam updated the diff for D61495: vmm: Route MMIO-to-MMIO MOVS by GPA.

Address review

Fri, Nov 13, 3:55 PM
hayzam requested review of D61499: bhyve: Preserve string I/O registers on errors before progress.
Fri, Nov 13, 6:22 AM
hayzam requested review of D61498: bhyve: Preserve upper RAX bits for byte and word input.
Fri, Nov 13, 6:14 AM
hayzam requested review of D61495: vmm: Route MMIO-to-MMIO MOVS by GPA.
Fri, Nov 13, 4:12 AM

Sun, Nov 8

hayzam updated the test plan for D61316: vmm: Mask CPUID leaf 0x80000008 EBX on Intel hosts as well.
Sun, Nov 8, 6:49 AM
hayzam requested review of D61316: vmm: Mask CPUID leaf 0x80000008 EBX on Intel hosts as well.
Sun, Nov 8, 6:48 AM

Fri, Nov 6

hayzam updated the test plan for D61277: camdd: Fix the worker startup race.
Fri, Nov 6, 11:02 PM
hayzam requested review of D61277: camdd: Fix the worker startup race.
Fri, Nov 6, 11:01 PM

Thu, Nov 5

hayzam added a reviewer for D61219: vmm: Avoid redundant Intel DEBUGCTL writes: kevans.
Thu, Nov 5, 7:39 PM
hayzam updated the summary of D61003: bhyve: Reject unsupported checkpoints before saving RAM.
Thu, Nov 5, 7:37 PM
hayzam updated the diff for D61003: bhyve: Reject unsupported checkpoints before saving RAM.

Add vm_can_snapshot_devices() and pci_can_snapshot() as suggested

Thu, Nov 5, 7:36 PM
hayzam updated the test plan for D61219: vmm: Avoid redundant Intel DEBUGCTL writes.
Thu, Nov 5, 10:16 AM
hayzam requested review of D61219: vmm: Avoid redundant Intel DEBUGCTL writes.
Thu, Nov 5, 10:15 AM
hayzam added reviewers for D61218: vmm: Avoid redundant AMD DEBUGCTL writes: bnovkov, kevans.
Thu, Nov 5, 6:14 AM
hayzam updated the test plan for D61218: vmm: Avoid redundant AMD DEBUGCTL writes.
Thu, Nov 5, 6:10 AM
hayzam updated the diff for D61218: vmm: Avoid redundant AMD DEBUGCTL writes.

Add sponsorship credit

Thu, Nov 5, 6:09 AM
hayzam updated the summary of D61218: vmm: Avoid redundant AMD DEBUGCTL writes.
Thu, Nov 5, 6:09 AM
hayzam requested review of D61218: vmm: Avoid redundant AMD DEBUGCTL writes.
Thu, Nov 5, 6:03 AM
hayzam added a reviewer for D61068: virtio_balloon: Add newer memory statistics tags: kevans.
Thu, Nov 5, 1:44 AM
hayzam added a reviewer for D61067: vmm: Add a shared guest-memory discard interface: kevans.
Thu, Nov 5, 1:44 AM