Implement sparse image support (no deps on libsparse)

linux-msm / qdl

BSD 3-Clause "New" or "Revised" License

225 stars 79 forks source link

Implement sparse image support (no deps on libsparse) #37

Open bmx666 opened 1 year ago

bmx666 commented 1 year ago

Hi @andersson, yep, it is another "sparse" version but without deps on libsparse. I've been testing it on several QCOM projects for over 2 years now, no problem.

parport0 commented 1 year ago

@bmx666 Hi,

Why is this PR a draft? Are there any flaws in this implementation?

bmx666 commented 1 year ago

Hi @parport0

Why is this PR a draft?

I'm waiting feedback from maintainer...

Are there any flaws in this implementation?

There are only two flaws:

upload speed can drop significantly if the image contains a lot of small chunks, compared to QPST / QFIL. Some kind of optimization is needed to merge and send many small fragments at once, because image may contain several small fragments following each other.
CHUNK_TYPE_FILL is not yet supported, but it can be easily implemented.

obbardc commented 1 month ago

Looks pretty good to me, thanks @bmx666 Slowdown is fine, I guess. But perhaps CHUNK_TYPE_FILL is more important ?

bmx666 commented 1 month ago

Looks pretty good to me, thanks @bmx666 Slowdown is fine, I guess. But perhaps CHUNK_TYPE_FILL is more important ?

Hi @obbardc , I uploaded an update to support CHUNK_TYPE_FILL a long time ago. Also during my research on Qualcomm tools, they replace the 4-byte chunks of fill values with one big CHUNK_TYPE_RAW. Anyway, I don't have access to any Qualcomm boards anymore, I couldn't confirm and verify this patch.