On Sun, Sep 10, 2023 at 01:46:01AM +0530, Mukesh Ojha wrote: > Hi All, > > This is to continuation from the conversation happened at v4 > > https://lore.kernel.org/lkml/632c5b97-4a91-c3e8-1e6c-33d6c4f6454f@xxxxxxxxxxx/ > > https://lore.kernel.org/lkml/695133e6-105f-de2a-5559-555cea0a0462@xxxxxxxxxxx/ > > We have put abstract on LPC on this topic as well as initiated a mail thread > with other SoC vendors but did not get much traction on it. > > https://lore.kernel.org/lkml/0199db00-1b1d-0c63-58ff-03efae02cb21@xxxxxxxxxxx/ > > We explored most of possiblity present in kernel to address this issue[1] but > solution like kdump/fadump does not seems safe/secure/performant from our > perspective. > > Hence, with this series we tried to make the minidump kernel driver, simple > and tied with pstore frontends, so that it collects the present available > frontends data like dmesg, ftrace, pmsg, ftrace., Also, we will be working > towards enhancing generic pstore to capture more debug data which will be > helpful for first hand of debugging that can benefit both other pstore users > as well as us as minidump users. > > One of the proposal made here, > https://lore.kernel.org/lkml/1683561060-2197-1-git-send-email-quic_mojha@xxxxxxxxxxx/ > > Looking forward for your comments. > > Thanks, > Mukesh > > [1] > Minidump is a best effort mechanism to collect useful and predefined data > for first level of debugging on end user devices running on Qualcomm SoCs. > It is built on the premise that System on Chip (SoC) or subsystem part of > SoC crashes, due to a range of hardware and software bugs. Hence, the > ability to collect accurate data is only a best-effort. The data collected > could be invalid or corrupted, data collection itself could fail, and so on. > > Qualcomm devices in engineering mode provides a mechanism for generating > full system ramdumps for post mortem debugging. But in some cases it's > however not feasible to capture the entire content of RAM. The minidump > mechanism provides the means for selecting which snippets should be > included in the ramdump. > > The core of SMEM based minidump feature is part of Qualcomm's boot > firmware code. It initializes shared memory (SMEM), which is a part of > DDR and allocates a small section of SMEM to minidump table i.e also > called global table of content (G-ToC). Each subsystem (APSS, ADSP, ...) > has their own table of segments to be included in the minidump and all > get their reference from G-ToC. Each segment/region has some details > like name, physical address and it's size etc. and it could be anywhere > scattered in the DDR. > > Existing upstream Qualcomm remoteproc driver[1] already supports SMEM > based minidump feature for remoteproc instances like ADSP, MODEM, ... > where predefined selective segments of subsystem region can be dumped > as part of coredump collection which generates smaller size artifacts > compared to complete coredump of subsystem on crash. > > [1] > https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/tree/drivers/remoteproc/qcom_common.c#n142 > > In addition to managing and querying the APSS minidump description, > the Linux driver maintains a ELF header in a segment. This segment > gets updated with section/program header whenever a new entry gets > registered. > > Changes in v5: > - On suggestion from Pavan.k, to have single function call for minidump collection > from remoteproc driver, separated the logic to have separate minidump file called > qcom_rproc_minidump.c and also renamed the function from qcom_minidump() to > qcom_rproc_minidump(); however, dropped his suggestion about rework on lazy deletion > during region unregister in this series, will pursue it in next series. > > - To simplify the minidump driver, removed the complication for frontend and different > backend from Greg suggestion, will pursue this once main driver gets mainlined. > > - Move the dynamic ramoops region allocation from Device tree approach to command line > approch with the introduction command line parsing and memblock reservation during > early boot up; Not added documentation about it yet, will add if it gets positive > response. > > - Exporting linux banner from kernel to make minidump build also as module, however, > minidump is a debug module and should be kernel built to get most debug information > from kernel. > > - Tried to address comments given on dload patch series. > > Changes in v4: https://lore.kernel.org/lkml/1687955688-20809-1-git-send-email-quic_mojha@xxxxxxxxxxx/ > - Redesigned the driver and divided the driver into front end and backend (smem) so > that any new backend can be attached easily to avoid code duplication. > - Patch reordering as per the driver and subsystem to easier review of the code. > - Removed minidump specific code from remoteproc to minidump smem based driver. > - Enabled the all the driver as modules. > - Address comments made on documentation and yaml and Device tree file [Krzysztof/Konrad] > - Address comments made qcom_pstore_minidump driver and given its Device tree > same set of properties as ramoops. [Luca/Kees] > - Added patch for MAINTAINER file. > - Include defconfig change as one patch as per [Krzysztof] suggestion. > - Tried to remove the redundant file scope variables from the module as per [Krzysztof] suggestion. > - Addressed comments made on dload mode patch v6 version > https://lore.kernel.org/lkml/1680076012-10785-1-git-send-email-quic_mojha@xxxxxxxxxxx/ > > Changes in v3: https://lore.kernel.org/lkml/1683133352-10046-1-git-send-email-quic_mojha@xxxxxxxxxxx/ > - Addressed most of the comments by Srini on v2 and refactored the minidump driver. > - Added platform device support > - Unregister region support. > - Added update region for clients. > - Added pending region support. > - Modified the documentation guide accordingly. > - Added qcom_pstore_ramdump client driver which happen to add ramoops platform > device and also registers ramoops region with minidump. > - Added download mode patch series with this minidump series. > https://lore.kernel.org/lkml/1680076012-10785-1-git-send-email-quic_mojha@xxxxxxxxxxx/ > > Changes in v2: https://lore.kernel.org/lkml/1679491817-2498-1-git-send-email-quic_mojha@xxxxxxxxxxx/ > - Addressed review comment made by [quic_tsoni/bmasney] to add documentation. > - Addressed comments made by [srinivas.kandagatla] > - Dropped pstore 6/6 from the last series, till i get conclusion to get pstore > region in minidump. > - Fixed issue reported by kernel test robot. > > Changes in v1: https://lore.kernel.org/lkml/1676978713-7394-1-git-send-email-quic_mojha@xxxxxxxxxxx/ > > Testing of the patches has been done on sm8450 target after enabling config like > CONFIG_PSTORE_RAM and CONFIG_PSTORE_CONSOLE and once the device boots up. > > echo mini > /sys/module/qcom_scm/parameters/download_mode > > Try crashing it via devmem2 0xf11c000(this is known to create xpu violation and > and put the device in download mode) on command prompt. > > Default storage type is set to via USB, so minidump would be downloaded with the > help of x86_64 machine (running PCAT tool) attached to Qualcomm device which has > backed minidump boot firmware support. > > This will make the device go to download mode and collect the minidump on to the > attached x86 machine running the Qualcomm PCAT tool(This comes as part Qualcomm > package manager kit). > > After that we will see a bunch of predefined registered region as binary blobs files > starts with md_* downloaded on the x86 machine on given location in PCAT tool from > the target device, more about this can be found in qualcomm minidump guide patch. > I tried to apply this series on top of 535a265d7f0dd50 (as suggested by `b4 am -l -g`), but it conflicts on patch [04/17]. Please specify the exact base commit or another series for which this series is based on. Thanks. -- An old man doll... just what I always wanted! - Clara
Attachment:
signature.asc
Description: PGP signature
