Hi, On 09/02/2020 04:48, Arthur D. wrote: > I suppose the real life scenario would be: > > 0. The OS runs on eMMC. > 1. The user opens his phone back cover and inserts MicroSD card. > 2. Kernel doesn't try to access the card until the cover is closed. This kinda makes sense - but assume the card is already in there, and the user wants to swap them? Then it doesn't help. > It seems wise to me: we don't want to damage user's data or hardware. > Accessing data on unstable medium can't be considered safe. > While the cover is open and user just inserted the microsd card everything > can happen - the phone can be dropped, may be unstable contact to microsd > while the user interacting it, etc. Maybe I am just used to the microsd port being exposed, but I have no other ARM device that doesn't have the microsd slot easily exposed. Many phones, cables, development boards, and they all just allow me to insert a microSD card, and it'll work. Clearly the manufacturers didn't have a similar worry there. > It seems to be inconvenient for us, while we run OS mostly from microsd. > But removing such behaviour from the kernel completely doesn't seem to > be a good idea generally. Vanilla kernel shouldn't be dedicated for > debugging purposes or edge cases (like running only from external MMC). > Running from embedded MMC should be common for most users in perspective. > > If I remember correctly, if the microsd is already mounted and the back > cover is open, nothing bad happens. It continues to work. You can unmount > the card and remove it safely. But you are not going to see the new card > insertion detected until you close the cover. I think the last paragraph might actually be key, because when you said that, my initial reaction was: "aha! that makes some sense". Still not sure if it warrants the hassle it currently is. And as it stands, we also disable this feature in Maemo Leste, so we clearly decided we don't want it either. ;-) > I'm not sure about the last paragraph. Please correct me if I'm wrong about > how it actually works. OK. If we don't want to remove this, can we at least somehow have a runtime warning? I spent quite some time searching my dmesg, trying to figure out why I couldn't even see the mmc interface that I needed. Cheers, Merlijn
Attachment:
signature.asc
Description: OpenPGP digital signature
