在 Android 上自定义 /system 挂载点

Question

我正在尝试将 Nexus 5 更改为使用 dm-verity 模块。为此，我需要：

1. 在引导期间在 /system挂载期间调用名为veritysetup的实用程序；
2. （此实用程序将创建一个名为 /dev/mapper/devname 的新设备）；
3. 让 /system 挂载到 /dev/mapper/devname，不是实际的分区。

看起来挂载点是在 fstab.hammerhead (device/lge/hammerhead) 上配置的。事实上，我有另一个模块通过更改这个文件来工作。所以这似乎是最简单的部分。

我知道有些文件系统是由 init.c (system/core/init) 挂载的，例如 /dev 和 /proc，而有些是由 init.hammerhead (system/core/rootdir) 挂载的，例如 /acct。但是我可以找到挂载 /system 的请求在哪里。

有什么想法吗？

Answer 1

这不是正确的做法。要使用 dm-verity，我们需要：

1. 使用“user”构建类型构建android（默认为“eng”）；
2. 更改 fstab 以在 fs_mgr_flags 中包含“verify”关键字。

因此，当 Android 时，vold 模块会找到 verity 标志并创建映射器设备。

如果系统映像具有签名哈希表和预先计算的哈希，并且根映像具有公钥，那么一切都应该正常工作。我们应该会发现系统分区挂载在 /dev/block/dm-0 而不是 /dev/block/platform/msm_sdcc.1/by-name/system 并且正在被读取时进行验证。

Answer 2

我有一个类似的询问，这里发布的答案是我找到的最相关的帖子，当然我的询问是相关的......

有谁知道生成 fstab 文件或将 fstab 文件放入 stage1 和 stage2 引导的代码在哪里，我认为某处存在错误或文件丢失。我可以轻松更改和更新内容。 AOSP 中的哪些文件实际上需要在构建之前进行编辑，以便在将 fstab 文件打包到 ramfs 阶段 1 或阶段 2 之前对其进行修改或更改。此外，我在哪里放置内核调试参数，如“pci=nocrs”？

*注意：此处接受的答案是根据 aosp 源文档的正确答案，但它遗漏了与源文档相同的有价值信息。在运行构建之前需要将文件放在树中的什么位置我可以看到输出确实有带有几个版本的 fstab 的供应商文件夹，但我不确定它在启动映像中的位置。

我想在 qemu kvm 中使用墨鱼构建启动它。 完整的 kernel.log 文件在这里：https://pastebin.com/9MKFKeyN 下面有趣的部分：

[    1.702829] init: Switching root to '/first_stage_ramdisk'
[    1.703038] init: [libfs_mgr]ReadFstabFromDt(): failed to read fstab from dt
[    1.703574] init: Using Android DT directory /proc/device-tree/firmware/android/
[    1.712562] init: [libfs_mgr]Invalid ext4 superblock on '/dev/block/by-name/metadata'
[    1.712864] traps: init[1] trap invalid opcode ip:41a6eb sp:7ffff59c93e0 error:0 in init[2f7000+200000]
[    1.713197] init: InitFatalReboot: signal 4
[    1.713574] init: #00 pc 000000000031b86b  /init
[    1.713626] init: Reboot ending, jumping to kernel
[    1.713718] kvm: exiting hardware virtualization
[    1.788726] reboot: Restarting system with command 'bootloader'
[    1.788872] reboot: machine restart

截至 2022 年 4 月 2 日的最新 AOSP 版本

构建道具 aosp/out/target/product/vsoc_x86_64/ramdisk/system/etc/ramdisk/build.prop

####################################
# from generate-common-build-props
# These properties identify this partition image.
####################################
ro.product.bootimage.brand=generic
ro.product.bootimage.device=vsoc_x86_64
ro.product.bootimage.manufacturer=Google
ro.product.bootimage.model=Cuttlefish x86_64 tv
ro.product.bootimage.name=aosp_cf_x86_64_tv
ro.bootimage.build.date=Sat Apr  2 12:06:36 CDT 2022
ro.bootimage.build.date.utc=1648919196
ro.bootimage.build.fingerprint=generic/aosp_cf_x86_64_tv/vsoc_x86_64:Tiramisu/AOSP.MASTER/me04021206:userdebug/test-keys
ro.bootimage.build.id=AOSP.MASTER
ro.bootimage.build.tags=test-keys
ro.bootimage.build.type=userdebug
ro.bootimage.build.version.incremental=eng.me.20220402.120943
ro.bootimage.build.version.release=12
ro.bootimage.build.version.release_or_codename=Tiramisu
ro.bootimage.build.version.sdk=32
# end of file

fstab.ext4 /genvol/aosp/out/target/product/vsoc_x86_64/recovery/root/first_stage_ramdisk/fstab.ext4

# Non-dynamic, boot critical partitions
/dev/block/by-name/boot /boot emmc defaults recoveryonly,slotselect,first_stage_mount,avb=boot
/dev/block/by-name/init_boot /init_boot emmc defaults recoveryonly,slotselect,first_stage_mount,avb=init_boot
/dev/block/by-name/vendor_boot /vendor_boot emmc defaults recoveryonly,slotselect
system /system erofs ro wait,logical,first_stage_mount,slotselect,avb=vbmeta_system,avb_keys=/avb
system /system ext4 noatime,ro,errors=panic wait,logical,first_stage_mount,slotselect,avb=vbmeta_system,avb_keys=/avb
# Add all non-dynamic partitions except system, after this comment
/dev/block/by-name/userdata /data ext4 nodev,noatime,nosuid,errors=panic latemount,wait,check,quota,formattable,fileencryption=aes-256-xts:aes-256-cts,keydirectory=/metadata/vold/metadata_encryption,checkpoint=block
/dev/block/by-name/metadata /metadata ext4 nodev,noatime,nosuid,errors=panic wait,formattable,first_stage_mount,check
/dev/block/by-name/misc /misc emmc defaults defaults
# Add all dynamic partitions except system, after this comment
odm /odm erofs ro wait,logical,first_stage_mount,slotselect,avb
odm /odm ext4 noatime,ro,errors=panic wait,logical,first_stage_mount,slotselect,avb
product /product erofs ro wait,logical,first_stage_mount,slotselect,avb
product /product ext4 noatime,ro,errors=panic wait,logical,first_stage_mount,slotselect,avb
system_ext /system_ext erofs ro wait,logical,first_stage_mount,slotselect,avb=vbmeta_system
system_ext /system_ext ext4 noatime,ro,errors=panic wait,logical,first_stage_mount,slotselect,avb=vbmeta_system
vendor /vendor erofs ro wait,logical,first_stage_mount,slotselect,avb=vbmeta
vendor /vendor ext4 noatime,ro,errors=panic wait,logical,first_stage_mount,slotselect,avb=vbmeta
vendor_dlkm /vendor_dlkm erofs ro wait,logical,first_stage_mount,slotselect,avb
vendor_dlkm /vendor_dlkm ext4 noatime,ro,errors=panic wait,logical,first_stage_mount,slotselect,avb
odm_dlkm /odm_dlkm erofs ro wait,logical,first_stage_mount,slotselect,avb
odm_dlkm /odm_dlkm ext4 noatime,ro,errors=panic wait,logical,first_stage_mount,slotselect,avb
system_dlkm /system_dlkm erofs ro wait,logical,first_stage_mount,slotselect,avb=vbmeta
system_dlkm /system_dlkm ext4 noatime,ro,errors=panic wait,logical,first_stage_mount,slotselect,avb=vbmeta
# ZRAM, SD-Card and virtiofs shares
/dev/block/zram0 none swap defaults zramsize=75%
/dev/block/vdc1 /sdcard vfat defaults recoveryonly
/devices/*/block/vdc auto auto defaults voldmanaged=sdcard1:auto,encryptable=userdata
shared /mnt/vendor/shared virtiofs nosuid,nodev,noatime nofail

如果有帮助，我还有完整的内核日志和其他信息，Vtd 和 Vtx 正在工作，并且所有 pci 都显示在 IOMMU 组中。

acloud create --local-instance 1 --local-image 告诉我如何在失败时开始故障排除，只有在您手动构建证书文件后，因为 instance1 的 acloud 设置的那部分无法完成，我发现了那个错误，但尚未修复它的来源，这很简单 - 手动创建证书允许它继续，但在引导中的 switchroot 失败。所有硬件和qemu都表现良好。

无法添加完整 kernel.log 的一些兴趣点

Waiting for AVD(s) to boot up ...stop_cvd I 04-02 17:33:50 137759 137759 main.cc:162] Successfully stopped device cvd-1: 0.0.0.0:6520
Fail! (453s)
Total time:  (453s)


Device summary:
Fail in:
Cannot create cuttlefish instance: Device did not boot within 450 secs. Stderr:
The following files contain useful debugging information:
  Serial console is disabled; use -console=true to enable it.
  Kernel log: /tmp/acloud_cvd_temp/local-instance-1/cuttlefish_runtime/instances/cvd-1/kernel.log
  Logcat output: /tmp/acloud_cvd_temp/local-instance-1/cuttlefish_runtime/instances/cvd-1/logs/logcat
  Launcher log: /tmp/acloud_cvd_temp/local-instance-1/cuttlefish_runtime/instances/cvd-1/logs/launcher.log
  Instance configuration: /tmp/acloud_cvd_temp/local-instance-1/cuttlefish_runtime/instances/cvd-1/cuttlefish_config.json
  Instance environment: /tmp/acloud_cvd_temp/local-instance-1/.cuttlefish.sh
Failed to read a complete exit code, read 0 bytes only instead of the expected 4
VIRTUAL_DEVICE_BOOT_FAILED
launch_cvd E 04-02 17:33:50 115944 115944 main.cc:252] run_cvd returned 10
For more detail: /tmp/acloud_cvd_temp/local-instance-1/cuttlefish_runtime/launcher.log

Encountered the following errors:
Cannot create cuttlefish instance: Device did not boot within 450 secs. Stderr:
The following files contain useful debugging information:
  Serial console is disabled; use -console=true to enable it.
  Kernel log: /tmp/acloud_cvd_temp/local-instance-1/cuttlefish_runtime/instances/cvd-1/kernel.log
  Logcat output: /tmp/acloud_cvd_temp/local-instance-1/cuttlefish_runtime/instances/cvd-1/logs/logcat
  Launcher log: /tmp/acloud_cvd_temp/local-instance-1/cuttlefish_runtime/instances/cvd-1/logs/launcher.log
  Instance configuration: /tmp/acloud_cvd_temp/local-instance-1/cuttlefish_runtime/instances/cvd-1/cuttlefish_config.json
  Instance environment: /tmp/acloud_cvd_temp/local-instance-1/.cuttlefish.sh
Failed to read a complete exit code, read 0 bytes only instead of the expected 4
VIRTUAL_DEVICE_BOOT_FAILED
launch_cvd E 04-02 17:33:50 115944 115944 main.cc:252] run_cvd returned 10
For more detail: /tmp/acloud_cvd_temp/local-instance-1/cuttlefish_runtime/launcher.log

If you have any question or need acloud team support, please feel free to contact us by email at buganizer-system+419709@google.com.


Line 227 of kernel log: [    0.318334] PCI: Using host bridge windows from ACPI; if necessary, use "pci=nocrs" and report a bug